6. DISCUSSION

A discussion of our photometry and a quantitative analysis of UV/optical galaxy morphologies will be presented in later papers. Here, we summarize the more important trends regarding the UV morphologies of galaxies which have emerged from our data and related studies.

1. UV morphologies are usually significantly different from visible band morphologies. This is a consequence of spatially inhomogeneous stellar populations. Differences are most pronounced for systems in the middle range of Hubble types, Sa to Sc. Objects at the extremes of the Hubble sequence (ellipticals or very late disk systems and irregulars) tend to have more homogeneous populations and therefore less dramatic wavelength-dependent morphology.

2. Although elliptical galaxies and large spiral bulges are fainter in the far-UV than the optical bands by factors of ~ 6-60 (in flux per unit wavelength), nonetheless they are much brighter than expected from the cool ~ 1 M turnoff stars in the dominant population. Their optical and UV axis ratios are similar, though they are usually much more compact in the UV. They exhibit smooth UV profiles, with none of the clumpiness normally associated with recent massive star formation. (Exceptions are NGC 1275, recently affected by a cooling flow or interaction, and the nonthermal jet in NGC 4486.) There is little UV evidence for dust in our sample of normal ellipticals at 3" resolution (though lanes are present in the peculiar objects NGC 1275 and 1316 and are often found at HST resolution in elliptical cores). Because of the absence of dust, the intergalactic ionizing radiation field of old stellar populations may be competitive with that of young populations in more dusty disk and irregular galaxies.

3. The ellipticals and spiral bulges exhibit large far-UV/optical color gradients in which the colors become redder outward (with the exception of NGC 221, where the gradient is strong but reversed). The gradients are larger than any at optical/IR wavelengths. The UV light is evidently extremely sensitive to the characteristics of its parent population. The FUV profiles of some of the early-type systems (e.g., NGC 221 and 1399) are well fitted by de Vaucouleurs functions, which are characteristic of spheroids at optical wavelengths. However, Ohl et al. (1998) find that the inner FUV profiles of some E/S0 galaxies are more consistent with an exponential function. Although exponentials are normally associated with disks, the FUV contours are generally consistent in shape and orientation with optical band isophotes, and the three-dimensional FUV light distributions are therefore unlikely to be genuinely disklike. Because of the large color gradients, it is not necessarily expected that the UV profiles of systems which are true spheroids at optical wavelengths would be closely de Vaucouleurs in shape.

4. The best evidence is that the far-UV light in most elliptical galaxies and large bulges is produced by extreme horizontal branch stars and their descendants in the dominant, low-mass, metal-rich population (O'Connell 1999 and references therein). The UV output of such objects is very sensitive to envelope properties and hence age, helium and metal abundance, and giant branch mass loss. It is a remarkable circumstance, if this interpretation is correct, that relatively crude UV observations can in principal determine envelope masses for stars in distant galaxies with a precision of a few 0.01 M .

5. The range in central (UV-optical) colors for early-type galaxies is very large, as first established by IUE spectra (e.g., Burstein et al. 1988). We find unusually red (MUV-R) values, after extinction corrections, for the early-type galaxies in the Perseus cluster. These may be metal-rich systems which lack the extreme HB component.

6. The cool stars in the large bulges of Sa and Sb spirals fade in the UV while the hot OB stars in their disks brighten such that their Hubble classifications (if these were attempted in the UV) would become significantly later. In the far-UV, early-type spirals can appear as peculiar, ringlike systems. At optical wavelengths, such structures would be associated with a short-lived, nonequilibrium situation, not with the stable configuration we know them to be.

7. The mean azimuthally averaged profiles of galaxy disks are approximately exponential in the UV but show considerably more structure than at optical wavelengths. In spiral disks with well organized, large amplitude spiral structure, the UV-bright structures closely outline the spiral pattern. In most cases, however, the UV appearance of disks can be considerably more fragmented and chaotic than at optical wavelengths, and the UV Hubble types are significantly later. UV suppression of bulge light in large-bulge spirals permits study of population changes in their inner disks, which appear to be dramatic in the case of M81.

8. The luminous AGNs in our sample range considerably in their contribution to the total UV light. In Mrk 335 the AGN is completely dominant; in NGC 4151 it contributes 80%; while in NGC 1068, 1275, and 4486 the contribution is only 10%-20%. A number of systems have unusual UV-bright structures in their inner disks, including rings (NGC 1317, 4321) and compact knots (NGC 1068). Such structures, and starburst nuclei (e.g., NGC 2993 here and M83 in Kuchinski et al. 2000), could easily dominate the UV light in high-redshift analogs.

9. Because of the often complex structural changes between the optical and UV bands, there is no simple relation between the half-light radii in these bands. However, in the most regular Sb-Sc spiral disks with good photometric coverage (NGC 628 and 3031), the UV light is more extended than the optical continuum light.

10. Although none are illustrated in this paper, the cool bars of barred spirals become less prominent in the UV as well (Bohlin et al. 1983; Kuchinski et al. 2000), and this effect extends even to fairly late-type systems such as Magellanic irregulars (Page & Carruthers 1981; Smith, Cornett, & Hill 1987).

11. Dust in normal spiral disks does not dominate UV morphologies, despite the high selective UV extinction characteristic of Milky Way and Magellanic Cloud dust grains. The overall distributions of far-UV light and H light for those systems where both are available are quite similar even though A(1500Å) / A(H) ~ 3.2. Such ready escape of far-UV photons from normal spiral disks was not anticipated. Quantitative analysis of UV spectra of starburst galaxies shows that the UV continuum often suffers less extinction than predicted from optical band extinction estimates, e.g., based on H / H ratios (Fanelli et al. 1988; Calzetti et al. 1994; Mas-Hesse & Kunth 1999). This is evidently mainly a consequence of the complex dust/star/gas geometries of star-forming regions. The emergent UV spectrum will, of course, always be weighted toward regions of lower extinction. But it appears that the heaviest extinction is confined to thin layers and the immediate vicinity of young H II complexes and that the UV light emerges from thicker star distributions, regions evacuated of dust by photodestruction or winds, or by virtue of strong clumpiness in dust concentrations (Fanelli et al. 1988; Calzetti et al. 1994; Calzetti 1997; Gordon, Calzetti, & Witt 1997; Allen et al. 1997). Only in cases like M82 and NGC 2146, where the dust layers appear dynamically disrupted, or the massive star formation is deeply embedded, does dust appear to be a major factor in UV morphology. A tentative conclusion from the available information is that dust will infrequently conceal the UV light of starburst activity for periods longer than ~ 50 Myr in the absence of continuing dynamical disturbances. This implies that some high-redshift starburst systems could be UV-bright.

12. Dust is expected to be more important in edge-on galaxies, and some of these can be very faint in the UV (e.g., NGC 891 here). However, other edge-on systems have high UV surface brightness, for instance UGC 6697 in this paper or NGC 4631 in Smith et al. (2000). Since the dust content of the outer disks of galaxies is expected to be low because of the lower metallicities there, one might expect many edge-on systems to be reasonably bright in the UV.

13. A significant fraction of the far-UV light in spiral disks is diffuse rather than being closely concentrated to obvious star-forming regions, and this appears to vary from object to object. The far-UV (1500 Å) light is in general more diffuse than is H. The diffuse component could arise from migration of UV-bright stars away from their birth sites during their ~ 50 Myr lifetimes, in situ star formation in low-density regions, or scattered light from dust grains. In the case of the Large Magellanic Cloud, UIT observations determined that about 70% of the detected 37,000 OB stars fell outside the boundaries of well-defined H II regions (Parker et al. 1998), consistent with diffuse component estimates from deeper, narrow-field HST observations (Brosch et al. 1999). The UV-bright plume of M82 indicates that dust scattering can be very important in some cases; similar instances of scattering in galactic winds might be detectable in high-redshift starburst systems.