2. GALAXY MORPHOLOGY AND ITS REDSHIFT EVOLUTION

With ground-based seeing, the study of galaxy morphology was restricted to redshifts of no more than a few tenths. The advent of HST, and its resolution of ~ 0.1", has revolutionized this field. Results from projects such as the HST Medium Deep Survey (MDS, Griffiths et al. 1994a, b) have shown that at faint magnitudes (I_AB > 21) an increasing fraction of galaxies do not conform to the traditional categories (e.g., Glazebrook et al. 1995; Driver et al. 1995). The first Hubble Deep Field (Williams et al. 1996) dramatically pushed this study to even lower fluxes, tracing sub-L* galaxies to high redshift. The optical images of the HDF show that by I_AB 24, the conventional Hubble sequence no longer provides an adequate description of many or most galactic systems (Abraham et al. 1996; Driver et al. 1998). Indeed, at higher redshifts we may be seeing new classes of galaxy emerge with no local counterpart, such as the 'chain galaxies' (Section 3 and Cowie, Hu & Songaila 1995) and 'tadpoles' (van den Bergh et al. 1996).

Some of these faint sources are intrinsically under-luminous peculiar galaxies at modest redshift. However, the median redshift has risen to z 1 for a limiting magnitude of I_AB = 26 (Lanzetta, Fernández-Soto & Yahil 1997). Hence, in the faint magnitude régime, band-shifting effects become important: the optical passbands sample shorter rest-frame wavelengths in galaxies at the higher redshifts, and large "morphological k-corrections" can arise (e.g., Odehahn et al. 1996). At z 1, the appearance in the observed optical is dominated by regions of recent star formation, luminous in the rest-frame ultraviolet on account of the massive, short-lived OB stars. Indeed, Colley et al. (1996) suggest that the observed peak in the two-point angular correlation function of the optical HDF at 0".3 is due to mis-classifying multiple compact star-forming regions within larger high-redshift galaxies as separate systems, exacerbated by the cosmological (1 + z)^-4 bolometric surface-brightness dimming which boosts the contrast between the compact star-forming knots and the more diffuse host galaxy.

2.1 High-Resolution Imaging in the Near-Infrared

The rest-optical is a far better tracer of the dynamical mass of a galaxy than the ultraviolet. This suggests a strategy of high-resolution imaging in the near-infrared; the V- and R-bands in the rest-frame of a z 1 galaxy are well approximated by the J- and H-passbands, and multi-colour imaging out to the H-band can trace the rest-frame B-band morphology of galaxies as far as z 3. However, until recently there has been no high-resolution infrared data set which reaches a limiting flux comparable to the optical HDF.

The Instrument Development Team (IDT) of the HST NICMOS camera (Thompson et al. 1998) have imaged an area of the northern HDF to unprecedented depth in the near-infrared, observing for 49 orbits in each of the F110W and F160W filters (centered at 1.1 µm and 1.6 µm and similar to the ground-based J- and H-bands). The widest-field NIC 3 camera was used to survey a ~ 1 arcmin² portion of the HDF. A detailed description of the observations and data reduction are given by Thompson et al. (1999). Once we correct for different resolutions of NIC 3 and WFPC 2 (through "PSF matching"), we can use the spatially-resolved colours to study different stellar populations and/or dust-reddening within a galaxy (see Figs. 2 & 5).

2.2 The Transformation of Spiral Galaxies with Wavelength

One of the most visually striking differences between the optical and near-infrared HDF images are spiral galaxies at moderately-high redshift (z ~ 1). At NICMOS wavelengths (the rest-optical), many of these are clearly classic spirals, and therefore dynamically-evolved stable systems which certainly should not fall under the banner of morphological peculiars. However, as illustrated in Fig. 1, moving to the rest-UV shifts the classification toward a much later Hubble type - i.e., becoming more irregular (Bunker, Spinrad & Thompson 1999). In extreme cases, the galaxy appearance is such a strong function of wavelength that some systems which resemble small groups of tidally-interacting sub-galactic clumps in the WFPC 2 optical images are only unveiled as nucleated spirals by the infrared observations. A classic example is the galaxy HDF 4-474.0 at z = 1.059 (Cohen et al. 1996) which is totally dominated by an off-centre star forming H II region in the U- and B-images, but transforms into a 'grand design' face-on spiral in the near-infrared (Fig. 1a). Spiral bulges are dominated by cool giants, and so brighten at the redder wavelengths; in the case of HDF 4-378 (at an estimated photometric redshift of z = 1.20, Fernández-Soto, Lanzetta & Yahil 1999) the bulge is totally absent from the observed optical passbands, but dominates the infrared light (Fig. 1b). This is reminiscent of the far-UV 1500 Å imaging with UIT of the local spiral, M81, presented in O'Connell (1997).

Figure 1. Spiral galaxies at z 1, showing the great change in apparent morphology going from the optical (the rest-ultraviolet, where the appearance is irregular) to the near-infrared, where their true spiral nature is revealed. In the case of HDF 4-474 (left), the WFPC 2 images are dominated by a star forming knot, and for HDF 4-378 (right) the older/redder population of the bulge is only visible at infrared wavelengths.

Figure 2. Stellar population fits to two spatially resolved regions of the z 1 spiral HDF 4-474 (see Fig. 1a), using the latest version of the Bruzual & Charlot (1993) models. The bulge (left panel) is clearly very much older than the star-forming H II region in one of the spiral arms (right panel).

From the optical HDF, there also appears to be strong redshift evolution in the relative fraction of galactic bars. Indeed, van den Bergh et al. (1996) report just one barred spiral in the whole of HDF-North. More recently, Abraham et al. (1999) have found similar evolution in the WFPC 2 images of HDF-South (Williams et al. 1998), with a marked decline at z > 0.5 in the proportion of barred spirals in both fields. If this is a truly evolutionary effect, then it has great significance for the physics of disk formation. However, once again the effects of large morphological k-corrections at higher-redshifts makes the case for evolution inferred from the apparent decline of barred spirals at faint optical magnitudes less clear cut. Bars are dominated by older stellar populations, with similar colors to bulges (de Vaucouleurs 1961), and so are prominent at redder wavelengths. In the rest ultraviolet, the star forming regions in the disk will typically dominate the light, and a spiral which would be identified as being barred when viewed in the rest optical may be (mis-)classified as unbarred at shorter wavelengths. Examination of the IDT-NICMOS images reveals bars in the near-infrared which are undetected in the WFPC 2 images (e.g., Fig. 3b at z 1); hence, claims of evolution in the frequency of galactic bars based on optical data alone should be treated with some caution.

Figure 3. The left panel shows the only optically-selected barred spiral in HDF-North (van den Bergh et al. 1996), and this seems to be through chance alignment of a swath of young stars with the approximate axis of the true bar. The galactic bar in the spiral displayed in the right panel is only recognizable at infrared-wavelengths - at its redshift of z 1, the optical wavebands only sample the rest-ultraviolet, where the older & redder bulge/bar stellar populations are not prominent.

2.3 The Redshift Evolution in the Fraction of Truly Peculiar Systems

Using the six wavebands from the WFPC 2 and IDT-NICMOS imaging of the Hubble Deep Field, we have compared galaxy morphology at the same rest-frame wavelengths. Where available, we use the spectroscopically-measured redshifts (from Cohen et al. 1996 unless otherwise noted). Where no published spectroscopic redshift exists, we adopt the photometric redshift estimate of Fernández-Soto, Lanzetta & Yahil (1999). Figure 3 of Bunker (1999) shows the rest-frame B-band of all the galaxies in the IDT-NICMOS field brighter than I_AB = 25, which extends out to z 3.

Down to I_AB 25.5 (the brightest 100 galaxies in IDT-NICMOS field), only about 1/6 of galaxies change their appearance greatly between the WFPC 2 and NICMOS images - these have large morphological k-corrections. Of the remaining number, about half of the galaxies retain the same morphology in all wavebands (above the redshifted Lyman break) and are "true peculiars". Hence, the increased fraction of unusually-shaped systems at faint optical magnitudes is largely due to evolution rather than simply band-shifting effects. The remaining third of galaxies are too compact for changes in morphology to be ascertained (limited by the NIC 3 PSF, which has a FWHM of 0.25 arcsec), and this fraction increases greatly at magnitudes fainter than I_AB = 25. For most cosmologies, the higher-redshift systems are on average more compact, once allowance has been made for the fact that the higher-redshift systems are intrinsically more luminous in this apparent-magnitude limited sample.