1.1. H II Regions in the Context of the ISM
The original two-phase model of the interstellar medium was the H II / H I region dichotomy introduced by Strömgren (1939). He showed that photoionized gas near hot stars is segregated into physically distinct volumes, separated from their neutral environment by sharp boundaries. The study of ionized (formerly "gaseous") nebulae differs in many ways from that of most of the other components of the ISM, for both historical and physical reasons. Because H II regions are the only form of interstellar material which emits strongly in the optical spectral region, there is a much longer and richer history of observations and theory for them than for the other thermal phases of interstellar matter. (The discovery of nebular emission lines dates back to the mid-1860's.) In contrast, most of our information about the other phases depends on recently developed technologies such as radio, infrared, and X-ray astronomy.
Optical observations of H II regions provide fairly complete information about their elemental composition. From their spectra, abundances relative to hydrogen can be estimated for nearly all of the most common elements, particularly He, N, O, Ne, Ar, and S. (Note that oxygen alone constitutes nearly 50% by mass of the elements heavier than helium.) Furthermore, ionized nebulae are remarkably efficient machines for converting ultraviolet continuum energy from OB stars, originally diluted over wide bandpasses, into a few narrow, intense, optically-thin emission lines. The intrinsic emissivities of these lines are easy to calculate in principle, although they are sensitive to the local thermodynamic state of the gas (ne and Te). On the other hand, the thermal parameters can also be determined from the spectra, using diagnostic line-intensity ratios. In this way, H II regions can be used to measure element abundances in the (present-day) gas of distant galaxies. The sample of extragalactic H II regions studied so far has metal abundances ranging from about .02 to several times solar. This is a useful complement to studies of our own Galaxy, which contains no severely metal-deficient H II regions (except for a handful of planetary nebulae formed by stars of the halo population). In contrast, for many H II regions in the outskirts of late-type spirals and in some dwarf irregular galaxies, the process of metal enrichment by stellar nucleosynthesis is still in its early stages, providing a "window" on the early chemical evolution of galaxies. The flip side of this coin is that these low-metallicity H II regions are also presumed to have experienced only a small degree of alteration in their helium abundances due to stellar activity. Therefore, their present He / H ratios should be nearly the same as the primordial value, providing valuable tests for cosmological theories.
1.2. Environments and Systematics
Although a comprehensive review of all the properties of extragalactic H II regions is beyond the scope of this paper, a brief discussion is needed to provide a context for the subject of abundances. The various "categories" of extragalactic H II regions are essentially lists of their environments. These include: (1) disk H II regions in spiral and irregular galaxies; (2) gassy dwarf irregular galaxies with spectra which are heavily dominated by H II regions; and (3) nuclear and near-nuclear regions sometimes called "starburst" or "hotspot" H II regions (e.g. Kennicutt, Keel, and Blaha 1989). In the present review I will concentrate on the first two categories, for which the best abundance data are available. Regions in the third group tend to have relatively strong stellar continua and to be fairly metal-rich, which make it to difficult to obtain accurate measurements of the emission lines from which abundances are determined. On the other hand, members of the first two categories are universally regarded as members of the same family. H II regions in nearby galaxies have been well-catalogued; atlases are available for the LMC, SMC, and a large number of other galaxies (Hodge and Wright 1967, 1977; Hodge and Kennicutt 1983). The star-forming dwarf irregulars are usually found by spectroscopic surveys for emission-line galaxies (see, for example, the review by Kinman 1984).
The statistical properties of the H II region populations in spiral and irregular galaxies were most recently addressed by Kennicutt (1988) and Kennicutt, Edgar, and Hodge (1989). They find that late-type galaxies have both intrinsically higher-luminosity first-ranked H II regions, and larger total numbers of H II regions after normalization by galaxy size, than do early-type spirals. Within a galaxy, the differential luminosity function of the H II regions is roughly power-law, N L-2 ± 0.5, although some low-luminosity irregulars have an exceptional supergiant complex, and Sa-Sb galaxies are deficient in luminous regions. While the positive correlation between the luminosity of the brightest H II region and that of the parent galaxy can be understood as chiefly a sample-size effect, the dependence on morphological type is a real and separate factor. Typical large galaxies contain hundreds of optically detectable H II regions. We note here that, of all the regions detected and cataloged in H or H, it is usually the nearest and the most luminous ("giant") H II regions for which abundances are derived.
The dimensions of these "giant" extragalactic H II regions are typically 100-200 pc. They are ionized by clusters of 101-4 OB stars and contain 103-5 M of ionized gas. Some of the best-studied regions are the 30 Dor complex in the LMC, NGC 604 in M33, and NGC 5461 and 5471 in M101. Certain selection effects must be present. Necessarily poorer spatial resolution contributes to a tendency to identify larger regions in more distant galaxies. This effect is illustrated by Israel, Goss, and Allen (1975), who compare large-beam radio measurements with optical images of the same H II regions in M101; at better resolution these regions break up into groups or chains of smaller clumps. Likewise, H II regions in dwarf irregulars are also found to have complex structure when closely examined (e.g. Hodge, Lee, and Kennicutt 1989; Davidson, Kinman, and Friedman 1989). In more distant galaxies, we will always be looking at more heterogeneous volumes; for example, a typical aperture size (4") for spectrophotometric studies corresponds to 1 pc at 50 kpc (the LMC) and 2 kpc at 100 Mpc.
1.3. Internal Structure
The morphology of many giant extragalactic H II regions can be characterized to first order as a "core-halo" structure, on the basis of both optical and radio-continuum data (see, for example, the multi-wavelength study of NGC 604 by Israel et al. 1982). The cores are composed of dense material, often in several distinct clumps, close to the ionizing stars. The diffuse, lower-density envelopes are presumably ionized by photons escaping from the inner regions and represent the radiation-bounded edges of the Strömgren volume. Most giant extragalactic H II regions are believed to be essentially radiation-bounded (e.g. McCall, Rybski, and Shields 1985). In addition, the denser regions themselves are inhomogeneous, as seen in the recent detailed studies of NGC 5471 by Skillman (1985), and of NGC 604 by Diaz et al. (1987). That there are also inhomogeneities on smaller spatial scales is shown by the discrepancy between (rms) ne values derived from recombination emission and local values determined from density-sensitive line ratios. The dense clumps are embedded in a much lower-density medium, with typical clump volume filling factors of .01 - .1 (e.g. Kennicutt 1984; McCall, Rybski, and Shields 1985). The interclump material is often treated as a vacuum in nebular models, because it does not contribute significantly to the optical emission lines.
A good deal of recent work has focused on the velocity structure of the emission lines. Giant extragalactic H II regions display supersonic velocities, which appear to correlate with H luminosity. Terlevich and Melnick (1981) interpret the line-widths as virial and therefore usable for determining the local gravitational field; they also find a secondary dependence on metallicity. An alternative interpretation of the origin of the line-widths is that they are a result of stellar winds from the exciting stars, and possibly also from embedded supernova remnants (e.g. Dopita 1981; Skillman 1985). For nearby regions, it is possible to actually identify the stars which may be responsible for driving the high-velocity gas, as in a recent kinematic model of NGC 604 by Clayton (1988).
1.4. Ionizing Clusters
Luminous extragalactic H II regions are ionized by OB associations. For nearby regions, the members of the stellar cluster can be distinguished individually and HR diagrams constructed (see papers in De Loore, Willis, and Laskarides 1986). The nebular ionization structure and emitted spectrum will evolve as the cluster ages and the UV radiation field diminishes and softens. Exploratory models of this kind have been calculated (e.g. Melnick, Terlevich, and Eggleton 1985; Kennicutt and Chu 1988). The process has been inverted, using the nebular spectrum to infer the age of the star cluster or "burst" (Lequeux et al. 1981; Copetti, Pastoriza, and Dottori 1985). A more controversial issue is whether the stellar initial mass function varies with metallicity, as proposed by Terlevich (1986). Over the last decade it has become apparent that Wolf-Rayet stars are often present in extragalactic H II regions. Wolf-Rayet features have been seen in M 33 by D'Odorico and Rosa (1981) and Conti and Massey (1981), and in many other regions as well (see Rosa and D'Odorico 1986 for a recent review and further references.) The frequency of Wolf-Rayet stars is higher for higher-metallicity regions (Maeder, Lequeux, and Azzopardi 1980), a point to which we return below. Wolf-Rayet stars are important in our context because they furnish metal-rich outflows which are capable of altering the chemical composition of their gaseous environment (Chiosi and Maeder 1986). Giant H II regions are also known hosts of Type II supernovae (Richter and Rosa 1984). Winds and supernovae from the massive stars can contaminate (or enrich) the local gas in H II regions in He, C, O, and other species. Evidence for such local enrichments has been sought and perhaps seen in some regions (Skillman 1985; Pagel, Terlevich, and Melnick 1986).