2. BACKGROUND AND MOTIVATION FOR ULTRAVIOLET IMAGING OF GALAXIES

Changes in the two-dimensional light distribution of galaxies with wavelength can provide new and unique perspectives on their structures and evolution. The spectral region of interest here lies between the Lyman cutoff of the interstellar medium at 912 Å and the cutoff of the terrestrial atmosphere near 3200 Å. This window contains information on the character of stellar populations, dust grains, interstellar gas, and AGNs which is largely independent of that in the familiar optical bands. In this section we discuss the potential utility of the UV in understanding galaxy evolution and progress to date in exploiting these opportunities.

2.1. Ultraviolet Probes of Galaxy Astrophysics

2.1.1. Ages and Metal Abundances of Stellar Populations

The UV has the highest sensitivity of any spectral region to stellar temperature and metal abundance, implying that it is especially valuable as a means of characterizing stellar populations, current star formation rates (SFRs), and star formation histories. Stars with surface temperatures above ~ 10,000 K (e.g., main-sequence stars with masses 3 M) are brighter in the UV than at longer wavelengths (Fanelli et al. 1992). Consequently, UV imaging or spectroscopy of star-forming galaxies permits direct detection of the massive stars responsible for most ionization, photodissociation, kinetic energy input, and element synthesis.

Figure 1 illustrates predicted UV-IR spectral energy distributions of stellar populations over timescales up to 3000 Myr. The strong time evolution of the UV compared to longer wavelengths is the reason for its utility in determining population ages. The sensitivity of the UV to stellar properties extends even to the cool ~ 1 M stars near the main-sequence turnoff in the oldest model shown in Figure 1. These solar-type stars dominate the mid-UV (2000-3200 Å) light in this model, and chemical composition as well as age influences the spectrum. The concentration of strong metallic absorption features is responsible for much of the short-wavelength structure in this model. The abundance sensitivity of selected UV spectral regions is discussed by Fanelli et al. (1992).

Figure 1. Synthetic spectral energy distributions of single generation stellar populations having Salpeter IMFs and solar abundances for ages 3-3000 Myr from Bruzual & Charlot (1993), showing the rapid evolution in UV amplitude and shape. The shape of the near-IR spectrum ( > 7000) is much less sensitive to age. Interstellar emission lines are not modeled here but would be absent in all except the 3 Myr SED.

In old, quiescent systems such as elliptical galaxies and spiral bulges, the UV offers a second major probe of stellar populations. Most old systems have been found to contain a very hot, low-mass stellar component with T_eff > 15,000 K which dominates the far-UV light. This probably consists of stars with very thin envelopes on the extreme horizontal branch and subsequent advanced evolutionary phases (reviewed in Greggio & Renzini 1990 and O'Connell 1999). Their UV output is predicted to be very sensitive to their envelope masses and compositions. Overall, UV spectra are powerful age and abundance diagnostics for both young and old populations.

2.1.2. Star Formation Histories for t < 1 Gyr

The UV offers a unique probe of the star formation history of galaxies on intermediate timescales of 10-1000 Myr. By "history" we mean the star formation rate as a function of time, SFR(t). The most widely used methods for estimating the recent star formation rate, SFR(t₀), involve optical emission lines such as H, radio continuum emission from hot gas or relativistic electrons, and far-infrared or submillimeter continuum emission from dust grains (e.g., Kennicutt 1998). Both emission lines and free-free thermal radio continuum depend on photoionization from massive stars and therefore reflect activity only over the past ~ 5 Myr, after which photoionization rapidly decreases. This period represents only 0.05% of the star-forming lifetime of a typical galaxy. Nonthermal radio emission powered by supernova-driven relativistic electrons is a useful index of massive star formation over the past few tens of Myr (Condon 1992) but is influenced by the character of the surrounding interstellar medium. Infrared ( > 10 µm) and submillimeter thermal emission from dust grains is likewise strongly influenced by the nonstellar environment and has intrinsically poor time resolution, since grains can be heated by photons from stars of all ages.

These conventional methods for estimating recent SFRs are based on the indirect effects of massive stars, involving the downconversion of UV photons by surrounding media, and have either restricted or ill-defined age sensitivity. By contrast, the vacuum UV offers a direct measure of the light from the massive star populations. Extinction by dust is often cited as a serious obstacle to using direct FUV observations to infer star formation rates. However, the photoionizing UV continuum ( 912 Å) which drives line and free-free emission is possibly yet more sensitive to extinction, while there are fewer empirical constraints on its nature. The actual effects of dust on the emergent UV light are smaller than naively expected (see Section 2.1.3 and 6). All of these methods are comparably affected by uncertainties in extinction.

The timescales which can be probed by observations of the 1200-3200 Å continuum range from less than 10 Myr to 1 Gyr. This critical interval is well sampled neither by the methods described above nor by the optical region (3200-9000 Å), which is influenced by the star formation history on longer timescales (more than a few Gyr). It is the "gap" which Gallagher, Hunter, & Tutukov (1984), for instance, were compelled to omit in their landmark study of spiral galaxy histories derived from H fluxes, B-band fluxes, and total masses.

An example of the additional information on galaxy star formation histories to be gained from UV continuum imaging is shown in Figure 2. This is a map comparing the H and far-UV morphologies of the well-known Sc galaxy M51. There are clearly large variations in the far-UV to H ratio. The ~ 10-50 Myr old populations (FUV-bright) are usually spread farther downstream from the putative density wave than the ~ 5 Myr old, H -bright populations, but the pattern is not entirely symmetrical. Diffuse far-UV light tends to "fill in" the spiral arms between intense concentrations of H light. There is a hint of multiple FUV wavelets, with feathery extensions inclined in pitch angle to the main spiral pattern. The UIT data for M51 are discussed further by Kuchinski et al. (2000). A similar comparison map based on a lower resolution UV image from the FOCA program (see Section 2.3) was published by Petit et al. (1996).

Figure 2. H-FUV difference map of the face-on Sc galaxy M51. The image shows the logarithmic difference between a continuum-subtracted ground-based H image and a UIT far-UV (1500 Å) continuum image taken by UIT during the Astro-2 mission. The map contrasts regions of current star formation (5 Myr), which are bright in H and appear as white/light gray in the image, with regions active during the era 5-50 Myr in the past, which are relatively brighter in the FUV continuum and are represented by dark gray/black in the image.

The sensitivity of different wavebands to a galaxy's star formation history is discussed further in the form of "history weighting functions" in Appendix A.

2.1.3. Cold Interstellar Material

The UV offers high sensitivity to interstellar dust and regions of concentrated cold material (i.e., potential star formation sites). Before the advent of UV observations of galaxies, it was widely assumed that this would actually be a serious disadvantage because the Galactic extinction law (e.g., Cardelli, Clayton, & Mathis 1989) yields A(UV) / A(V) > 2.5, implying that the UV light of typical disk galaxies might be strongly suppressed. However, as is amply demonstrated by the images in this atlas and spectroscopic studies (e.g., Calzetti, Kinney, & Storchi-Bergmann 1994), dust does not dominate the UV morphology of most galaxies.

The UV may ultimately prove to be a valuable tracer of quiescent, cold, molecular material. Because the albedo of dust is high in the UV, cool interstellar clouds far from star-forming regions can be detected by scattered light, as in the case of the faint gaseous outer arms of M101 (Donas et al. 1981; Stecher et al. 1982) or the outer halo of NGC 1068 (Neff et al. 1994). UV imaging can also detect H₂ in photodissociation regions directly by virtue of its fluorescence bands in the 1550-1650 Å region (e.g., Witt et al. 1989; Martin, Hurwitz, & Bowyer 1990). Although such regions occupy only a small fraction of the total volume of a typical molecular cloud, nonetheless the direct detection of H₂ has, in principle, considerable advantages over methods involving trace constituents like radio-emitting CO (see Allen et al. 1997 and references therein).

2.1.4. Hot Interstellar Material

The UV contains uniquely important emission line probes of interstellar gas in the T ~ 10⁵-10⁶ K regime, including C IV (1550), N V (1241), and O VI (1035). These spectral diagnostics have been extensively exploited in absorption-line spectroscopic studies of our galaxy. Little has been done to date using imaging, though C IV images of supernova remnants have been published (e.g., the Cygnus Loop, Cornett et al. 1992; N49 in the LMC, Hill et al. 1995c).

2.1.5. Nuclear Structures

The nuclei of galaxies are often UV-bright. The optical-UV energy distributions of AGNs and associated nonthermal jets are relatively flat, and their contrast against the stellar background is usually better in the UV than in the optical band. Activity has been detected by UV imaging in a number of nearby galaxies (e.g., Maoz et al. 1995; Renzini et al. 1995; Barth et al. 1998). Spiral nuclei often contain starburst cores or ringlike structures, which are again more prominent in the UV (e.g., in M83, Bohlin et al. 1983; NGC 1068, Neff et al. 1994). The leverage of the UV in isolating such hot sources is especially important in spiral bulges and E galaxies, where the cool star background is often overwhelming at optical wavelengths.

2.1.6. Low Surface Brightness Systems

A minimum in the natural night sky background occurs at 1600-2400 Å. This is the deepest window in the UV-optical-IR spectrum and permits detection of extremely low surface brightness objects, perhaps up to 100,000 times fainter than the ground-based sky (O'Connell 1987; Waller et al. 1995). Applications include the study of faint circumgalactic star-forming regions in nearby galaxies and faint blue galaxies at redshifts up to ~ 1 (Martin et al. 1997; Treyer et al. 1998) and detection of low surface brightness disk systems (O'Neil et al. 1996).

2.2. Applications to High-Redshift Galaxies

UV imaging of nearby galaxies is relevant to galaxies at high redshift for two reasons. First, as just described, the rest-frame UV continuum is a robust tracer of star formation and is measurable to very high redshifts (z 10) with optical/IR instruments. For instance, the 2800 Å rest-frame continuum has been used to estimate the cosmic star formation density at z ~ 0.5-4 for the Canada-France Redshift Survey, Hubble Deep Field, and other surveys (Pei & Fall 1995; Lilly et al. 1995; Madau et al. 1996; Steidel et al. 1999), leading to the conclusion that gas processing occurred at a relatively constant rate for z ~ 1-4 but has precipitously declined since z ~ 1.

Second, observations of high-z galaxies are preferentially made in the rest-frame UV. This is particularly true for ground-based telescopes, where the rapidly increasing night sky brightness for wavelengths above 7000 Å, and thermal emission beyond 2 µm, seriously compromises observations in the near infrared. Because of the strong changes in galaxy appearance with wavelength, as illustrated in this atlas, there is a large "morphological k-correction" which must be quantified in order to distinguish genuine evolutionary effects from simple bandshifting.

High-redshift studies are also strongly influenced by reduced spatial resolution and by surface brightness selection. The latter is a very serious problem for z 1 because I ~ I₀(1 + z)^-n, where I₀ is the surface brightness in the rest-frame and n = 3 or 5 for monochromatic surface brightnesses per unit frequency or wavelength, respectively; n = 4 for bolometric surface brightnesses. Figure 3 illustrates these effects using a far-UV image of the luminous, nearby Sc spiral M101. Bohlin et al. (1991) and Giavalisco et al. (1996) describe methods of creating such simulations from rest-frame UV data.

Figure 3. Left panel: A far-UV (1500 Å) image of the luminous Sc galaxy M101 obtained by UIT during the Astro-2 mission. A 5' bar is shown for scale. Right panel: A simulation of a galaxy with the same structure but 10 times higher surface brightness at a redshift z = 1.5 as observed by the Keck 10 m telescope in a 10 hr exposure with 0".5 FWHM seeing against a sky background of 22.5 mag arcsec-2. A 2" bar is shown for scale. The simulation is not easily recognizable as a normal spiral galaxy. Its asymmetries are emphasized; it appears distorted and fragmented. High surface brightness star-forming associations in its disk have taken on the appearance of nearby "companions."

It is possible to explore bandshifting effects using multicolor (e.g., B and R) optical images to extrapolate the spectral energy distribution to the rest-frame UV on a pixel-by-pixel basis. This has been done using ground-based data (e.g., Abraham 1997; Abraham, Freedman, & Madore 1997; Brinchmann et al. 1998) and moderate-redshift Hubble Space Telescope (HST) data (Bouwens, Broadhurst, & Silk 1998). These studies, as well as those based on HST/NICMOS observations in the rest-frame optical bands (e.g., Bunker 1999, Corbin et al. 2000), conclude that the peculiarities in shape and size distributions found in the deep HST surveys considerably exceed the effects of bandshifting. While this is probably a robust result, these extrapolation methods do not accurately capture the range of rest-frame UV spectra found in real galaxies and are not suitable for making detailed comparisons with the local universe. The reason is that there is considerable scatter in UV colors of nearby galaxies at any given optical color. This is true even in the classic (U-B, B-V) diagram (e.g., Larson & Tinsley 1978), but it is much more pronounced in the rest-frame UV, where, for instance, Donas, Milliard, & Laget (1995) found a 3 mag range in (UV-b) colors at a given (b-r) color in a faint galaxy sample. This UV/optical decoupling is confirmed in spectroscopy of UV-selected samples by Martin et al. (1997) and Treyer et al. (1998). The implication is that the true evolutionary history of galaxies on timescales more recent than a few Gyr can be rather different from that inferred from optical data.

Fiducial photometric and imaging studies of nearby galaxies in the rest-frame UV are needed in order to calibrate these selection and morphological effects and to improve our understanding of the astrophysical drivers of the rest-frame UV luminosity, particularly the influence of dust and the history of star formation.

2.3. UV Imaging of Galaxies

Extragalactic UV astronomy to date has been largely based on spectroscopy, usually with small entrance apertures (1"-20") centered on galaxy nuclei (e.g., IUE, HST/FOS, HST/GHRS). Several hundred objects have also been photometered in broad bands with large apertures. Early photometric surveys were performed by OAO and ANS; the most recent large-scale study was by FAUST (Deharveng et al. 1994). The total number of UV spectroscopic and photometric observations of galaxies still far exceeds the number of imaging observations (an inversion of the historical development of optical extragalactic astronomy). Brosch (1999) has recently reviewed the available results of UV surveys of galaxies. No all-sky UV survey faint enough to include galaxies has yet been conducted, but this will be remedied by the GALEX mission (Martin et al. 1999).

The first UV image (defined as having many resolution elements over the area of interest) of another galaxy was obtained by the NRL Apollo S201 camera from the lunar surface in 1972 (Page & Carruthers 1981). This was of the Large Magellanic Cloud in the 1250-1600 Å band and dramatically demonstrated a strong wavelength-dependent morphology. Its remarkably fragmented appearance in the UV is entirely different from its familiar barlike shape in the optical continuum.

Subsequent progress in UV imaging up to 1990 was relatively slow (reviewed in O'Connell 1991). Since 1990, we have accumulated a sample of vacuum UV images of about 200 galaxies, principally from three instruments:

1. The Hubble Space Telescope Faint Object Camera (HST/FOC). Because of its better rejection of red leaks (i.e., residual response to low-energy photons outside the primary UV bands) the HST/FOC has been more extensively used for UV imaging than the HST/Wide Field Planetary Cameras. Its main limitation is a small field of view. FOC produced the first extensive compilation of galaxy UV images in an atlas of the nuclei (22" fields) of 110 nearby galaxies at 2300 Å with 0".05 resolution (Maoz et al. 1996). Many of the objects exhibit UV-bright nuclear structures, including a number of star-forming rings and five cases of unusually bright LINER active nuclei (Maoz et al. 1995). Far-UV (<2000 Å) imaging even with FOC is compromised by red leaks, but the newer HST/Space Telescope Imaging Spectrograph camera provides pure UV imaging over 25" fields (e.g., Brown et al. 2000).
2. The SCAP/FOCA balloon-borne telescope of the Marseille and Geneva groups. SCAP/FOCA operates in a narrow atmospheric window near 2000 Å with a large field of view of up to 2°.3 and resolution of ~ 15" (Deharveng et al. 1980; Milliard, Donas, & Laget 1991; Donas et al. 1995). By virtue of a fast f/ratio and long integration times, it can reach relatively low UV surface brightnesses despite its narrow band. A number of bright galaxies have been studied morphologically and photometrically with SCAP/FOCA. Donas et al. (1987) summarized integrated UV photometry for a sample of 149 spiral and irregular galaxies. A sample of fainter, mostly unresolved, UV-detected galaxies near the north galactic pole has been recently discussed by Donas et al. (1995) and Treyer et al. (1998). Fewer objects have been studied with resolved imaging; individual citations are given in Section 5.
3. The Ultraviolet Imaging Telescope (UIT), from which the UV images in this atlas were obtained. UIT has a field of view of 40' and spatial resolution of 3" and was intended to provide a good match to the performance of typical ground-based telescopes on large angular diameter, nearby galaxies. UIT is described in detail in the next section.

Because of technical difficulties in achieving high reflectivity optics shortward of 1100 Å and in rejecting the very bright geocoronal Ly line at 1216 Å from exposures centered at shorter wavelengths, both the HST and UIT imaging cameras work primarily at wavelengths longer than 1216 Å.