4.3.4. Si/O and Fe/O
Interstellar dust grains play an important role in the ISM; they absorb ionizing radiation, absorb and scatter starlight, remove heavy elements from the gas phase, and are the reputed sites for the formation of molecules. One of the major uncertainties in our understanding of grains is their composition, and how the composition and abundance of dust vary with environment. Much of our knowledge of the composition of dust comes from the study of features in the interstellar extinction curve (such as the 2175 Å absorption bump), and from measurements of interstellar absorption lines. The absorption line measurements reveal that many heavy elements have ISM abundances that are smaller than their "cosmic" values (defined by the atomic abundances measured in the solar photosphere or meteorites), and thus it is inferred that the unseen portion is incorporated, or depleted, onto dust grains (see reviews by Jenkins 1987, 1989). In most cases, the depletions show a dependence on the average gas density along the line of sight. It is also possible to observe emission lines from ions of refractory elements in ionized nebulae, from which one can infer the properties of grains in the ionized gas and study how grains may be modified in the vicinity of hot stars.
Figure 12. Top: The abundance ratio log Si/C plotted against log O/H for our extragalactic HII region sample. Filled circles represent our FOS measurements. Open squares represent abundances in the Orion Nebula from IUE spectra. The open stars represent the average Si/C ratios from different studies of Galactic B stars and A supergiants (Kilian 1992, Kilian et al. 1994, Kilian-Montenbruck et al. 1994, Cunha & Lambert 1994, and Venn 1995) Bottom: Log Si/O vs. log O/H for our HII regions. Symbols are the same as above. (From Garnett et al. 1995b).
Figure 12 displays the variation of Si/O and Si/C versus O/H for a sample of HII regions with a range of metallicities (from Garnett et al. 1995b). Also shown are Si/O and Si/C in the Orion Nebula, the Sun (Grevesse & Noels 1993), and in solar neighborhood B stars and A supergiants. Figure 12 shows (1) Si/C in the HII regions declines with increasing O/H by about 0.5 dex over the range -4.8 < log O/H < -3.1, and (2) Si/O is remarkably constant over the same metallicity range. The weighted mean value for the seven objects is log Si/O = -1.59 ± 0.07. For comparison, log Si/O = -1.37 in the Sun. The decline in Si/C reflects the increasing C/O ratio with O/H discussed in Section 4.3.2.
Measurements of gas-phase abundances for Si from interstellar absorption lines in the Galactic ISM typically show that Si is mostly depleted onto grains, with the depletion dependent on the mean density of the absorbing clouds (Jenkins 1987; Sofia, Cardelli, & Savage 1994). Si should also be depleted in our HII regions, which are relatively dense gas associated with regions of star formation, and we attempt to use our Si/O measurements to estimate the depletion of Si in the ionized gas.
From Figure 12 it can be seen that the range of average values for Si/O determinations in Galactic stars implies an uncertainty of approximately 0.3 dex in the reference value. If we compare the average Si/O in our HII region sample with the averages for the different samples of Galactic stars, the implied depletions for Si lie between -0.4 dex and 0.0 dex, that is, 40-100% of the Si in the HII regions is in the gas phase. If we account for about 0.1 - 0.2 dex depletion of oxygen in the HII regions, the inferred Si depletions lie between -0.6 dex and -0.1 dex (assuming Si/O is constant with metallicity, Timmes, Woosley, & Weaver 1995).
There are few measurements of refractory element depletions in HII regions for comparison. Iron abundances have been measured in three Galactic HII regions: Orion (Osterbrock, Tran, & Veilleux 1992), M17 (Peimbert, Torres-Peimbert, & Dufour 1993), and M8 (Peimbert, Torres-Peimbert, & Ruiz 1992). These studies inferred Fe depletions of -0.5 to -1.2 dex. These depletions are consistent with our measured Si depletions, if the trend of Fe depletion as a function of Si depletion shown by Sofia et al. (1994) holds everywhere. However, only one object (Orion) has had both Si and Fe abundances measured; a larger sample of measurements of both Si and Fe in HII regions is needed. Thuan et al. (1995) observed Fe III emission in a number of low metallicity blue compact galaxies and derived iron abundances which showed essentially no depletion. However, these were based on a single iron line (with a potentially uncertain emission coefficient), and I would urge caution in interpreting these abundances at present.
The fact that we see an average Si depletion of only -0.3 dex in the observed HII regions suggests that the grains in the ionized gas have been modified. Draine & Salpeter (1978) showed that thermal sputtering of grains in a hot gas is very inefficient at temperatures below 105 K, so this process is unlikely to be significant in photoionized gas at 104 K. Therefore, it may be that shocks are modifying the grains within the HII region. These could arise from stellar winds from O and Wolf-Rayet stars impinging upon the ionized gas, or from supernovae. Direct evidence for shocks in some giant HII complexes comes from detections of supernova remnants associated with them (Skillman 1985, Chu & Kennicutt 1986) and of diffuse X-ray gas within giant HII complexes (Williams & Chu 1995). The low Si depletions we see in the HII regions suggest that grain modification may be a general phenomenon in HII regions. It should be noted, however, that grain destruction appears to be rather incomplete within the HII regions: although relatively low depletion factors are observed for Si and Fe in the HII regions compared to the diffuse ISM, the measurements indicate that most of the Fe (and presumably, other elements such as Ca and Mg) is still in grains.
Our measurements of Si depletions appear to be consistent with the results of Sofia et al. (1994), who found that the Si/Fe and Si/Mg ratios in dust cores were smaller than expected if most of the Fe and Mg in dust cores is in silicates. They inferred that the dust cores must be predominantly metal and metal-oxide grains, while Si resides mainly in grain mantles. The low Si depletions we measure within HII regions appear to support this picture; we suggest that grains within HII regions have their mantles eroded, releasing a large fraction of the Si into the gas phase, while Fe-Mg grain cores may survive largely intact. Additional measurements of the abundances of silicon and other refractory elements in a variety of HII region environments will be needed to substantiate this idea. Remember that grain destruction or erosion within HII regions has important consequences for photoionization modeling (see Section 2.2).