2.2. WFPC2 B-band results and discussion
2.2.1. The scale-length - magnitude relation
Fig. 1 shows the scale-length rhl (averaged over the BVI filters in which the object was detected to increase S/N) versus B-magnitude for both the W02 and the HDF galaxies. The three main galaxy classes resulting from the rest-frame ANN are indicated with different colors or shades, as are the formal 50% completeness limits for both fields computed for nearby RC3 galaxies of the same types (see O96). Also plotted are the best fit models to the local rhl vs. MB relations for RC3 mid-type spirals and ellipticals for a median rhl = 4.7 and 3.2 kpc, respectively, assuming model redshift distributions (cf. CC92, NW95a) and K-corrections (BC93, D95a & b) as a function of galaxy type (using q0 = 0.5; but in one case also for q0 = 0.01). Generally, rhl is smallest at a given flux for E/S0's. The median scale-length at B 27 mag is rhl 0".25-0".3, which corresponds to ~ 1.2-2.5 kpc for the redshift range z ~ 0.5-2.5. These values appear to be smaller than the characteristic scale-lengths of mid to late-type galaxies measured locally, as well as those measured with HST for z 0.8 (or B 23 mag; cf. Mu94). A possible explanation is that the faint galaxy population becomes progressively more dominated by lower luminosity and therefore smaller late-type objects at fainter fluxes (cf. D94, D95a, D95b & Section 3).
2.2.2. The color-magnitude diagram
The total B450-magnitude versus (B - R) color is shown for ANN each type in Fig. 2, where "wide R" (V + I)/2 was used to increase S/N in the colors. Fig. 2 becomes increasingly incomplete at B 26.0 mag for the bluest W02 objects, because of the red detection limit at R 26.5 mag, as indicated by the slanted dashed lines (the HDF limits are ~ 1.5 mag fainter due to its finer substepping and the rhl vs. B-mag relation of Fig. 1). Only objects detected in all three filters are plotted here, but the fraction of objects seen only in B450 (at B 27 mag) and not in V606 and/or I814 - as well as those seen only in V606 & I814 ( 26 mag) and not in B450 - is 5%. Of significance is the clear segregation between the early and the late type galaxies for B 27 mag, even though no color information was used by the ANN classifier. E/S0's are almost without exception the reddest galaxies at any flux level, and Sd/Irr+M's are generally blue, at least down to the formal detection limit. Mid-type spirals (Sabc's) have colors with large dispersion and about as blue as Sd/Irr's, as expected from the (slightly) different colors measured for these populations locally and the K-corrections for galaxies with ongoing star formation at z 1-2 (BC93). The increase in the relative fraction of late-type and merging galaxies (Sd/Irr+M) in Fig. 3 towards fainter B450 is quite remarkable. They exist essentially down to the very detection limit, despite their larger relative scale-lengths (Fig. 1). The redder objects-mostly classified as E/S0 galaxies do not increase as rapidly towards the completeness limits (B 27 mag) as the late-types. Since early-type galaxies do not generally have the largest scale-lengths (Fig. 1) nor the lowest SB, this is probably not due to large differences in their redshift distribution - and therefore due to a larger cosmological SB-dimming. Instead, it may suggest that the formation of early-type galaxies was largely complete by the median redshift corresponding to B 26-27 mag (probably z 1, see NW95a), so that their counts should indeed be converging for B 26 mag (see O96 and Fig. 3b).
2.2.3. The galaxy counts versus type for B 27 mag
Fig. 3 gives the differential B number counts for the different morphological types (E/S0's, Sabc's, Sd/Irr+M's) in the ~ 36 WFPC2 B450-fields discussed in 2.1.1. The plotted counts are based on mean types from the visually classified U, B, V, & I images (lines coded by colors or shades), from the I814 + V606 single-filter ANN types (data points), and the rest-frame ANN types (heavy solid lines). Following D95a, D95b & O96, the B number counts were modeled for the three main morphological types separately. The best fit non-evolving models in Figs. 3b-3d are based on the local LF's of Ma94 (long-dashed curve) and Lo92 (dotted curve). Fig. 3b shows that the B-band counts of E/S0's follow the predictions for passively or mildly evolving models for B 24 mag, and for B 24 mag the observed counts are at most 0.3 dex higher than these models. Fig. 3c shows that Sabc's are consistent with these models for B 24 mag, and are at most 0.5 dex higher for 24 B 27 mag. Hence, a scenario invoking strong luminosity evolution is not required for the early-type galaxies out to B 27 mag (I814 25 mag or z ~ 1), suggesting that their formation was largely complete by z ~ 1. From a spectroscopic survey for z 1, L95 deduce a similar lack of evolution for early-type galaxies as found in D95a and O96. The morphological HST studies can push this work now another 3 magnitudes fainter than can be done spectroscopically from the ground.
Fig. 3d shows that non-evolving models are inadequate to explain the high WFPC2 B number counts for the Sd/Irr+M population. The fraction of visually classified Mergers are plotted as blue or filled squares. The fraction of mergers rises strongly for B450 23 mag, with up to one third of the late-type systems possibly being mergers. For B450 22 mag, the number counts of the classical Sd/Irr galaxies from both the ANN and visual types are well above the nonevolving models. It is clear that the FBG excess is dominated by galaxies with late-types, with a non-negligible fraction of merging morphologies. A non-evolving dwarf-rich population fits the FBG counts quite well [steep LF with = -2.0 (short-dash-dot) in Fig. 3d], but is inconsistent with the redshift distributions observed for B 24.5 mag (cf. PD95), and therefore cannot be the only explanation. Consistent with the I814-band counts of D95a, the Sd/Irr+M B450-counts in Fig. 3d is best described by an evolving Sd/Irr+M population with a moderately steep local LF [starburst with Lum = 1.5 mag at z 0.5 and a local Marzke LF (long-dash-dot), with the high normalization of D95b], that underwent significant luminosity evolution since z 1 (cf. PD95). The late-type galaxy population therefore must have evolved strongly and/or be dwarf-dominated, and causes most of FBG excess.
In such a scenario, a non-negligible fraction of the FBG's with 25 B 29 could be at z 1-3. In Section 3, we argue that these objects may have been the reservoir of building blocks from which the luminous early-type galaxies were formed through repeated merging - a process that must have been largely completed by z 1 for the early-type galaxies (Fig. 3b & c), with perhaps some residual epoch-dependent merger rate (cf. B94). The fact that 20-35% of the galaxies classified as late-type by the ANN appear to be merging systems - with that fraction increasing at fainter magnitudes tends to support this possibility. With the completion of ~ 100 B-band parallel fields in Cycle 6-7, we hope to extend the B450 counts with sufficient statistics at the bright end, covering the entire range 18 B 28 mag, thereby significantly constraining the "LF normalization problem" (Ma94) and the evolution for each population separately.