Annu. Rev. Astron. Astrophys. 1997. 35: 503-556 Copyright © 1997 by Annual Reviews. All rights reserved

6.4. Halo Alpha Elements

The enhancement of elements in the halo has been confirmed by numerous studies, both in the field and the globular cluster system (e.g. Clegg et al 1981, Barbuy et al 1985, Luck & Bond 1985, François 1987, 1988, Gratton & Sneden 1988, 1991, Zhao & Magain 1990, Nissen et al 1994, Fuhrmann et al 1995, McWilliam et al 1995a, b).

Because -element yields are predicted to increase with increasing SN II progenitor mass (e.g. Woosley & Weaver 1995), the [ / Fe] ratio is sensitive to the IMF. Therefore it is interesting to know if the [ / Fe] ratios in the halo are constant with changing [Fe/H], if there is a slope to the [ / Fe] correlation with [Fe/H], or if there is a measurable dispersion at a given [Fe/H], which might indicate a change in the IMF.

Abundance studies of oxygen are frequently based on the weak [O I] forbidden lines at 6300 and 6363 Å for cool giants (e.g. Barbuy 1988) and the high excitation O I triplet lines at 7774 and 9263 Å for main-sequence stars (e.g. Tomkin et al 1992). Unfortunately, the O I lines have very high excitation potential, and the resulting abundances may be very sensitive to temperature uncertainties and non-LTE effects. Tomkin et al (1992) found [O/Fe] ~ +0.8 dex from the O I lines, with non-LTE calculations; but the strong temperature dependence suggests that oxygen abundances derived from the triplet lines are unreliable. On the other hand, the [O I] lines are very weak and frequently only the 6300-Å line can be measured; however, the [O I] results are considered more reliable than those from the O I lines because neither temperature or non-LTE effects are a problem. Recent oxygen abundances have been determined from OH lines in the UV by Bessell et al (1991), Nissen et al (1994), whereas Balachandran & Carney (1996) used near-infrared OH lines. Both methods offer the advantage that many lines can be measured without severe non-LTE problems; but the reduced flux in the UV result in lower S/N and less reliable results for the UV OH lines than for the near-infrared OH lines.

All of these methods provide information on the free oxygen (uncombined into molecules) in the stellar atmospheres. However, for the total oxygen abundance, carbon abundances must also be known in order to account for the oxygen atoms locked up in the CO molecule.

The scatter in measured [O/Fe] values has been large: For example, (Abia & Rebolo 1989) found [O/Fe] = +1.0 for stars near [Fe/H] = -2.0, based on the O I triplet at 7774 Å. This result is almost certainly too high, as shown by many investigations (e.g. Barbuy 1988, Bessell et al 1991, Spite & Spite 1991, Kraft et al 1992, Nissen et al 1994). King (1993) suggested that the Abia & Rebolo equivalent widths were too high by approximately 25%, which, when combined with a revision of the temperature scale by 200 K, resolves the differences between the abundance results for O I lines and other oxygen abundance indicators.

The low S/N OH line results of Bessell et al (1991), Nissen et al (1994) suggest that [O/Fe] = +0.5 to +0.6 dex in the interval [Fe/H] = -1 to -3.4. Bessell et al (1991) claimed that the halo [O/Fe] ratios continue the slope of the [O/Fe] relation with [Fe/H] seen in the disk, down to [Fe/H] = -1.7; below this point the halo [O/Fe] ratio is constant.

From high S/N ~ 150) spectra, Barbuy (1988) measured a mean [O/Fe] = +0.35 ± 0.15 from the 6300-Å [O I] line in 20 halo giants with metallicities in the range -2.5 [Fe/H] -0.5. Kraft et al (1992), Sneden et al (1991b) measured [O/Fe] for many globular cluster giants and 27 field giants from S/N ~ 150 spectra of the [O I] line. The field giants ranged in [Fe/H] from -1.3 to -2.8 with an average [O/Fe] = +0.34. For oxygen-rich giants in the globular clusters (those without envelope depletion of oxygen) measured by Kraft et al (1992) the mean [O/Fe] = 0.32.

Balachandran & Carney (1996) measured C and O abundances in a halo dwarf using high S/N spectra of near-infrared OH and CO lines; they also rederived abundances from published O I, [O I], and C I lines. In particular, the solar and stellar abundances were both computed from the same grid of model atmospheres with the same set of lines. Balachandran & Carney found [O/Fe] = +0.29 dex for this star and concluded that temperature corrections were not required to resolve differences between forbidden and high excitation O lines, as had been previously suggested by King (1993). The resolution of the high O abundance values of Abia & Rebolo (1989) was due to the use of a self-consistent solar and stellar model atmosphere grid.

The preferred results of Figure 3a indicate a trend of [O/Fe] with [Fe/H] that is flat between [Fe/H] -1 to -3, at [O/Fe] = +0.34 dex; the dispersion of 0.1 dex about this value is consistent with the measurement uncertainties. Thus, the oxygen abundances in the bulge are consistent with a constant IMF. In Figure 3b I show results from O I triplet lines and low S/N spectra of UV OH lines, which exhibit a large dispersion.

There have been many studies for Mg, Si, Ca, and Ti in halo field stars; some of the more recent examples include those of François (1986), Magain (1987, 1989), Gratton & Sneden (1987, 1988, 1991), Zhao & Magain (1990) Ryan et al (1991) Nissen et al (1994), Fuhrmann et al (1995), McWilliam et al (1995a), Pilachowski et al (1996). Not surprisingly, the [ / Fe] ratios from this list encompass a range of values; for calcium, the lowest measured mean ratio for halo stars is [Ca/Fe] = +0.18 (Gratton & Sneden 1987), and the highest is [Ca/Fe] = +0.47 (Magain 1989). Much of the scatter in the abundance ratios is probably due to systematic effects in the analysis of different researchers: For example Gratton & Sneden consistently find lower [ / Fe] ratios than Magain Zhao & Magain; the usual differences are approximately 0.15 dex. Taking straight average abundance ratios for all the above studies gives the following results: [Mg/Fe] = +0.36, [Si/Fe] = +0.38, [Ca/Fe] = +0.38, and [Ti/Fe] = +0.29, with typical 1 = 0.08. The average of all four species gives [ / Fe] = +0.35, with = 0.05 dex, which is very close to the adopted value for [O/Fe] of +0.34 dex. A conservative conclusion is that in the halo, the elements O, Mg, Si, Ca, and Ti all show an enhancement, relative to Fe, of +0.35 dex; alternatively, the full range of measured [ / Fe] ratios is well represented by +0.37 ± 0.08.

François (1987, 1988) measured sulfur in halo stars from extremely weak, high excitation S I lines near 8694 Å and found [S/Fe] = +0.6; given the difficulty associated with abundance measurement of such weak lines, this result is approximately consistent with the general -element overabundances.

A small slope in the halo relation between [ / Fe] abundance ratios and [Fe/H] may possibly be responsible for part of the dispersion in published abundance results. For example, the McWilliam et al (1995a, b) study has a lower mean metallicity than Gratton & Sneden's (1988, 1991) work, with some overlap; on average McWilliam et al (1995a, b) [ / Fe] ratios are ~ 0.1 dex higher. However, the six stars common to both studies show a mean difference [McWilliam et al minus Gratton & Sneden] for [Fe/H], [Mg/Fe], [Si/Fe], [Ca/Fe], and [Ti/Fe] of only 0.00, 0.06, -0.08, -0.03, and 0.12 dex, respectively. Plots of [Mg/Fe] and [Ca/Fe] by McWilliam et al 1995a, b) showing the comparison with Gratton & Sneden (1988, 1991) could be interpreted as evidence for increases in both ratios with declining [Fe/H].

Gratton (1994) combined the -abundance results of several studies and found small increases in [O/Fe], [Ti/Fe], and [Mg/Fe] with declining [Fe/H] in the interval [Fe/H] = -1 to -3; this was claimed to be consistent with an increased production of O, Ti, and Mg at the lowest metallicity by high mass SN. The slopes were also consistent with chemical evolution model predictions of Matteucci & François (1992). Subtle slopes can also be seen in the [Ca/Mg] and [Ti/Mg] results of McWilliam et al (1995a, b), which may indicate a slight decrease in Mg, or an increase in Ca and Ti abundances, at the lowest metallicity. If true, these subtle trends indicate that the halo IMF was not constant with time. However, caution is warranted here because the small gradients could easily be the result of systematic measurement errors. Contrary to the above finding, Nissen et al (1994) concluded from analysis of halo field stars that the halo IMF was constant with time in the interval -3.5 [Fe/H] -1.8.

Carney (1996) reviewed -element abundances in globular clusters and found no evidence for a decline in [O/Fe], [Si/Fe], or [Ti/Fe] from [Fe/H] = -2.2 to -0.6 dex. A hint of a decline in [Ca/Fe] with increasing metallicity was found, which might be real but could equally well signal an analysis problem. The conclusion was that there is a uniform enhancement of [/Fe] ~ +0.3 dex in the globular clusters, with no evidence of SN Ia nucleosynthesis products in the younger clusters. The ~ 3 × 10⁹ year dispersion in globular cluster ages implies that either the time scale for SN Ia is longer than ~ 3 × 10⁹ years or that the "old halo" and "disk" globular clusters do not share a common history; at least one of the classes presumably formed far from the Galaxy and was accreted at a later time. Also, no large changes in IMF occurred during the epoch of globular cluster formation.

In principle, it should be possible to estimate the mean mass of SN that occurred in the halo because SN nucleosynthesis predictions (e.g. Arnett 1991, Woosley & Weaver 1995) indicate that certain ratios (e.g. O/C and O/Mg) are sensitive to progenitor mass. Unfortunately the predictions of the two theoretical papers are not entirely consistent, which makes it difficult to constrain the IMF.

Establishing whether the all elements exhibit the same level of enhancement in metal-poor stars is important; if so, this would favor a scenario in which the -element trend is due simply to the addition of iron-peak elements, as suggested by (Tinsley 1979). In this regard, McWilliam et al (1995a, b), Nissen et al (1994) measured [Ti/Fe] values ~ 0.1 dex smaller than enhancements of other elements (Mg, Si, Ca); McWilliam et al (1995a, b) claimed that this may be evidence that some Ti is produced in SN Ia.

There is increasing evidence for depletion of elements abundances in some halo stars: Fuhrmann et al (1995) found [Mg/Fe] = -0.28 in BD +3 740. McWilliam et al (1995a, b) found two stars (CS22968-014 and CS22952-015) near [Fe/H] = -3.4 with [Mg/Fe] < 0.0. These Mg depletions could be primordial, or they could be due to operation of the MgAl cycle in these stars (e.g. Shetrone 1996a, b), although the MgAl cycle would suggest large enhancements of Al, which are not seen in McWilliam et al (1995a) stars.

Brown et al (1996) found low [ / Fe] ratios in two young globular clusters (Rup 106 and Pal 12), with [Fe/H] of -1.5 and -1.0, respectively. In Pal 12, [Mg/Fe], [Ca/Fe], and [Ti/Fe] are approximately solar (i.e. below the halo value); in Rup 106, [O/Fe] and [Mg/Fe] are roughly solar, but [Ca/Fe] and [Ti/Fe] ~ -0.2 dex. Recently, Carney (in preparation) found a metal-poor field star with [Fe/H] = -1.9 and subsolar [ / Fe] ratios. The low -element abundances, compared with the general halo, suggest that these two globular clusters and the field star formed from material with an unusually large fraction of SN Ia ejecta. One explanation is that star formation in the parent clouds proceeded over time scales longer than the time delay for SN Ia. Such an event could occur in low-mass clouds with relatively low star formation rates; because high-mass stars form much less frequently than low-mass stars, a fraction of clouds could be expected to escape SN II for long periods of time and thus permit enrichment by SN Ia. One is reminded of the Taurus molecular cloud, which is currently forming low-mass stars only. Another possibility is that the star and clusters were captured from a companion galaxy, like the LMC, which experienced chemical evolution over an interval of time longer than the characteristic SN Ia time scale.

Bazan et al (1996) found enhanced -element abundances for a number of "metal-rich" halo stars (compared to the mean halo [Fe/H] value of -1.6 dex); the sample ranges from [Fe/H] ~ -1 to 0.0, with a mean near -0.5 dex. If confirmed, this shows that the enhanced -element abundances are a characteristic of the halo as a population, regardless of metallicity. This is supported by the -element overabundances in Arcturus measured by Balachandran & Carney (1996). This underscores the fact that the knee in the [ / Fe] versus [Fe/H] diagram (e.g. Figures 1 and 3) simply represents the intersection of the -element trends for the halo and disk and does not indicate an evolutionary connection.

Timmes et al (1995) made predictions of Galactic abundance ratios, from H to Zn, based on a Simple chemical evolution model and theoretical nucleosynthesis yields for SN II (from Woosley & Weaver 1995), SN Ia (from Nomoto et al 1984, Thielemann et al 1986), and 1- to 8-M stars (Renzini & Voli 1981). The predicted trends with metallicity for O, Mg, Si, and Ca show reasonable agreement with observations; but the predictions for S lie below the observed [S/Fe] ratios and are just barely consistent with the 0.3-dex theoretical uncertainty. The predictions for Ti are by far the worst of all elements; the theoretical [Ti/Fe] ratio is almost 0.7 dex below the observed values near [Fe/H] ~ -2. It is clear that present SN nucleosynthesis calculations completely fail to account for the observed [Ti/Fe] ratios in the Galaxy; Ti is significantly enhanced in the bulge and halo, yet nucleosynthesis calculations suggest that it should scale with Fe. Thus, Ti provides an important constraint for SN nucleosynthesis theory.