Carbon abundances in H II regions have been difficult to determine with precision because the important ionization states (C+, C+2) have no strong forbidden lines in the optical spectrum. Only the UV spectrum shows collisionally excited lines from both C II and C III. A number of studies of carbon abundances in extragalactic H II regions were made with IUE (e.g., Dufour, Shields, & Talbot 1982; Peimbert, Peña, & Torres-Peimbert 1986; Dufour, Garnett & Shields 1988), but for the most part the IUE observations suffered from low signal/noise and uncertainties due to aperture mismatches between UV and optical spectra.
The higher UV sensitivity of HST offered greatly improved measurements of UV emission lines from [C II] and C III], plus the opportunity to scale the C lines directly to [O II] and O III] lines in the UV, tremendously reducing the uncertainties due to reddening corrections and errors in Te (Garnett et al. 1995a, 1999; Kobulnicky & Skillman 1998). The most recent data for C/O as a function of O/H in dwarf irregular and spiral galaxies from HST measurements are displayed in Figure 16. Some C (and O) is expected to be depleted onto interstellar dust grains. Sofia et al. (1997) showed that the gas-phase C abundance varies little with physical conditions in the local neutral ISM, suggesting a constant fraction of C in dust everywhere. They infer that C is depleted by about 0.2 dex. O should be depleted by no more than 0.1 dex everywhere (Mathis 1996). Thus, it is likely that our C/O values should all be increased by 0.1-0.2 dex, but we do not expect any systematic variation in the fractional depletions with metallicity.
Figure 16. C/O abundance ratios (by number) from spectroscopy of H II regions in spiral and irregular galaxies (Garnett et al. 1995a, 1997b, 1999). Open symbols: irregular galaxies; filled symbols: spiral galaxy H II regions.
Figure 16 shows a trend of steeply increasing C/O for log O/H > -4. This is in agreement with observations of C/O in disk stars in the Galaxy (Gustafsson et al. 1999). The C/O ratios in the most metal-poor galaxies are consistent with the predictions for massive star nucleosynthesis by Weaver & Woosley (1993; hereafter WW93) for their best estimate of the 12C(, )16O nuclear reaction rate factor. On the other hand, the amount of contamination by C from intermediate mass stars is poorly known in these galaxies.
The notable trend in Figure 16 is the apparent `secondary' behavior of C with respect to O, despite the fact that C (i.e., 12C) is primary. Tinsley (1979) demonstrated that such variations can be understood as the result of finite stellar lifetimes and delays in the ejection of elements from low- and intermediate-mass stars. If C is produced mainly in intermediate-mass stars, then the enrichment of C in the ISM is delayed with respect to O, which is produced in high-mass stars.
At the same time, C is also produced in high-mass stars, with a production yield that is fairly uncertain. In stars without mass loss, the relative yield of C with respect to O in massive stars is smaller than the solar ratio (WW93), which would demand that most C come from intermediate-mass stars. Maeder (1992), however, showed that stellar mass loss can affect the yields of C and O from massive stars. The effect of such mass loss is to remove He and C from the massive stars before they can be further processed into O. If the mass-loss rates depend on radiative opacity, and thus on metallicity, then the yields of C and O will depend on metallicity, with the C yield increasing with Z at the expense of O.
Figure 17 shows the data for the spiral galaxies M101 and NGC 2403 with the predictions of two sets of chemical evolution models overlaid. The left panels show a sequence of Galactic chemical evolution models using massive star nucleosynthesis models including stellar winds from Maeder (1992); the right panels shows two other Galactic chemical evolution models derived with massive star yields computed assuming no stellar mass loss. All of the models use the same intermediate-mass star yields. Although all of the models reproduce the O/H gradients reasonably well, only the models with Maeder yields seem able to reproduce the steep C/O gradients observed - with the caveat that these models were not tailored for the two galaxies in question. Comparison of solar neighborhood models with the observations of stars also tend to favor the nucleosynthesis models that take into account metallicity-dependent mass loss for massive stars (e.g., Prantzos et al. 1994; Carigi 2000).
Figure 17. C/O and O/H gradients in M101 (open squares) and NGC 2403 (filled circles) plotted vs. radius normalized to the disk scale length (Garnett et al. 1999). Left panels: Galactic chemical evolution models from Carigi 1996, using massive star yields from Maeder 1992. The curves show the O/H and C/O gradients at different times, from 0.5 Gyr (dotted curves) to 13 Gyr (solid curves). Right panels: Galactic chemical evolution models from Götz & Köppen 1992 (dashed line) and Mollá et al. 1997 (solid lines) using massive star yields with no stellar winds. All models use the same intermediate-mass star yields from Renzini & Voli 1981.
The big uncertainties in all of this revolve around the theoretical yields. Problem number one is the 12C(, )16O reaction rate, which is still highly uncertain (Hale 1998). Problem two is uncertainty in convective mixing. 16O is produced by captures onto 12C during helium burning. Mixing of fresh He into the convective zone can turn C into O rapidly (Arnett 1996, pp. 223-229). Finally, mass loss rates for stars in various evolutionary states and metallicities are also still uncertain. For intermediate-mass stars, differences in mixing and treatment of thermal pulses affect the C yields. The most recent models for nucleosynthesis in intermediate-mass stars still show large discrepancies in yields (Portinari et al. 1998; van den Hoek & Groenewegen 1997; Marigo et al. 1996, 1998). Until these problems are solved or we have empirically-derived C yields for stars of various masses, it will be difficult to reliably interpret the abundance trends.