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

8.2. Halo Heavy Elements

To summarize the principle result of this section: Heavy element abundances in the halo are characterized by a significant r-process component and a 300-fold dispersion in [heavy element/Fe] ratios below [Fe/H] = -2.5 at about a constant average value. The pattern of r-process nucleosynthesis in the halo heavy element abundances provides evidence that this dispersion reflects an inhomogeneous composition of the material from which the stars formed, and the observed [heavy element/Fe] values set the minimum [heavy element/Fe] range from SN II events. The decreasing dispersion with increasing metallicity is consistent with a gradual homogenization of low metallicity gas by the process of averaging yields from individual SN events.

Although present in the abundance results of Wallerstein et al (1963), Pagel (1968) was the first to recognize that very metal-poor halo stars show heavy element deficiencies. Evidence for a plateau with [heavy element/Fe]~ 0.0, followed by a systematic trend of decreasing [heavy element/Fe] below [Fe/H] ~ -2.5, was presented by Spite & Spite (1978, 1979), Luck & Bond (1981, 1985), Barbuy et al (1985), Gratton & Sneden (1988), Magain (1989), Zhao & Magain (1990, 1991). In particular, the results of Gilroy et al (1988), Magain (1989) showed that [Eu/Fe] and [Zr/Fe] are enhanced in the interval [Fe/H] = -1.5 to -2.5.

Perhaps the most accurate abundance results for the largest sample of heavy elements come from Gratton & Sneden (1994), with stars in the interval -3 [Fe/H] -0.5. The abundance ratios [M/Fe] for Sr, Y, Ba, La, Ce, and Nd lie approximately between [M/Fe] ~ 0.00 to ~ -0.1 dex, and Zr, Sm, Pr, Dy, and Eu are enhanced relative to the solar composition. The ~ 0.3-dex enhancement in [Eu/Fe] is notable because Eu is an almost pure r-process element, thought to be produced only in SN II; its enhancement resembles the ~ 0.3-dex enhancement observed for the elements in the halo, which are also thought to be produced only in SN II.

The first evidence for a heavy element abundance dispersion in halo stars was due to Griffin et al (1982), who found heavy element enhancements in HD 115444. Luck & Bond (1985) found several halo stars with Ba and Sr enhancements. Both studies suggested that stars with heavy element enhancements were population II barium stars. ⁽²⁾ Gilroy et al (1988) claimed an r-process pattern and a large abundance dispersion for the halo heavy elements, which was consistent with heavy element abundance scatter in the material from which the stars formed. Ryan & Norris (1991) also claimed a dispersion in heavy element abundances larger than the measurement errors.

However, Baraffe & Takahashi (1993) suggested that the scatter was due entirely to measurement errors, based on the large scatter in published heavy element abundances for individual stars. The accurate abundance measurements of Gratton & Sneden (1994) indicated a heavy element dispersion of less than 0.1 dex for their sample, consistent with their measurement uncertainties.

McWilliam et al (1995a, b) analyzed a large sample of extremely metal-poor halo stars from the survey of Beers et al (1992), in the metallicity range -4 [Fe/H] -2, and made reliable estimates of the measurement uncertainties. They found a decline in [heavy element/Fe] below [Fe/H] = -2.5, accompanied by a considerable scatter in [Sr/Fe] and [Ba/Fe] abundance ratios, with a range of 2.5 dex (see Figure 9); typical measurement uncertainties were ± 0.2 dex. This scatter does not conflict with the small dispersion found by Gratton & Sneden (1994) because the Gratton & Sneden sample included only two stars below [Fe/H] = -2.5.

Figure 9. Plots of [Sr/Fe] and [Ba/Fe] from (McWilliam et al 1995a; filled circles) and (Gratton & Sneden 1994, 1988; open squares). The error bars indicate 1 uncertainties on the McWilliam et al results. The general run of the [Sr/Fe] data indicates a downward trend below [Fe/H] = -2.5 with a dispersion of ~ 300-fold. The large crosses represent the average [Sr/Fe] and [Ba/Fe] ratios for the McWilliam et al sample taken for 0.5-dex bins. The [Ba/Fe] data are similar to the [Sr/Fe] trend, but there is a hint of a bifurcation. Note that the star at [Fe/H] = -2.36 and [Ba/Fe] = +2.67 represents CS 22898-027, a CH subgiant, which is contaminated by s-process material accreted from an evolved companion, and so was not included in the average.

Ryan et al (1996) analyzed additional extremely metal-poor stars and found a large scatter in heavy element abundances for metal-poor halo dwarfs and giants, consistent with a primordial abundance scatter.

The McWilliam et al results show that for both [Sr/Fe] and [Ba/Fe] the mean ratios are the same above and below [Fe/H] = -2.5. Although there are many more stars deficient in heavy elements than with overabundances, the few heavy element-rich stars cause the average [Sr/Fe] and [Ba/Fe] ratios to be near the solar value. This explains why early studies found a trend of declining heavy element/Fe ratios as [Fe/H] declined; there is no trend, only a lower envelope of the dispersion that is skewed to low [Sr/Fe] and [Ba/Fe] ratios. Small samples preferentially picked the low [M/Fe] ratio stars because they are more frequent than the stars with heavy element enhancements.

Pagel (1968), Truran (1981) noted that the near-solar value of the [heavy element/Fe] ratio in the halo implies that the formation time for these elements must be shorter than the lifetimes of stars that produce s-process elements. Truran concluded that the heavy elements in the halo were made in massive stars by the r-process.

Abundance studies of halo stars by Sneden & Parthasarathy (1983), Sneden & Pilachowski (1985), Gilroy et al (1988) indicated heavy element abundance patterns consistent with nucleosynthesis dominated by the r-process.

The [Ba/Eu] ratio is particularly sensitive to whether nucleosynthesis of the heavy elements occurred by the s-process or r-process. The ubiquitous subsolar [Ba/Eu] ratios in halo stars (e.g. Magain 1989, François 1991, Gratton & Sneden 1994, McWilliam et al 1995b) show that the halo must contain a larger fraction of r-process material than the solar composition (e.g. Spite 1992). Figure 10 shows a compilation of [Ba/Eu] and [La/Eu] ratios for field halo stars, with pure r-process and s-process values indicated. The subsolar ratios indicate a larger fraction of r-process material than in the Solar System material; however, some s-process contribution may be required.

Figure 10. (a) The trend of [Ba/Eu] versus [Fe/H] and (b) [La/Eu] versus [Fe/H]: field stars represented by open boxes (Gratton & Sneden 1994) and triangles (McWilliam et al 1995). Star symbols represent mean globular cluster values from Brown et al (1992), and filled pentagons indicate globular clusters from the data of Shetrone (1996a), Armosky et al (1994). The open triangle indicates the CH subgiant, CS 22898-027, which is contaminated by s-process material. Dashed lines indicate the observed solar system r-process ratio (Käppeler et al 1989) and an extreme s-process value from Malaney (1987).

Cowan et al (1996) measured abundances of the r-process peak elements Os and Pt from UV lines in the metal-poor halo giant HD126238 ([Fe/H] = -1.7). When combined with abundances based on optical spectra, the best-fit heavy element pattern contains 80% r-process and 20% Solar System mixture. This is consistent with the value of [La/Eu] = ~ -0.4 for halo stars in the range -2 [Fe/H] -1 seen in Figure 10.

Element abundances for the most heavy element rich star known, CS 22892-052, were measured by Sneden et al (1994), Cowan et al (1995), McWilliam et al (1995b), Sneden et al (1996). Figure 11 shows the heavy element abundance pattern in this star, for elements heavier than Ba, which is identical to the Solar System r-process pattern (Käppeler et al 1989). Based on the large r-process overabundance and low [Fe/H], these authors concluded that the heavy elements in CS 22892-052 are dominated by the nucleosynthesis products of a single SN event. This does not mean that all the elements in this star are dominated by SN nucleosynthesis from a single event, only the heavy elements.

Figure 11. The heavy element abundance pattern in star CS 22892-052, from Sneden et al (1996), scaled to the barium abundance. The line represents the observed Solar System r-process abundance pattern from Käppeler et al (1989). The excellent agreement suggests that nucleosynthesis was dominated by the r-process; the small scatter about the r-process line indicates that the error bars were overestimated.

Cowan et al (1995) showed that some s-process contribution would help the fit to the observed Sr and Y abundances in CS 22892-052, but the Zr abundances cannot be explained by the s-process; thus the r-process probably at least contributes to the Zr abundance in the Sr-Y-Zr peak. McWilliam et al (1995b) showed that the [Sr/Ba] ratio is approximately constant in halo stars, despite the factor of 300 range of barium abundance due to the r-process. Therefore, if Sr has a significant contribution from the s-process, then the s-process to r-process ratio must be roughly constant in the halo; an alternative is that the r-process is a dominant source of Sr in the halo.

Magain (1995) measured the abundances of barium isotopes in one halo star in order to find the relative contribution of r- and s-process nucleosynthesis, from a profile fit to the Ba II line at 4554 Å. The best-fit profile indicated a Solar System mixture of barium isotopes, contrary to that expected from r-process nucleosynthesis. To resolve the discrepancy with element abundance ratios it would be very useful to have Ba isotopic compositions for a larger sample, especially for stars with both strong and weak Ba II lines.

François (1996) also disputed the claimed r-process source of halo heavy elements, based on a plot of [Eu/H] versus [Ba/H], arguing against the break in slope of the Ba) versus (Eu) seen by Gilroy et al (1988). However, it is better to rely on diagnostic abundance ratios (like [Ba/Eu] and [La/Eu]) as a discriminant of the nuclear reactions involved, rather than on the presence or absence of a break in slope.

Sneden et al (1996) noted that total Ba abundances measured from strong Ba lines in halo stars depend upon the assumed r- and s-process fractions. This effect may result in a downward revision of many previously reported Ba abundances by ~ 0.1 to 0.2 dex and bring earlier measurements of the [Ba/Eu] ratio closer to the pure r-process value. Therefore, when considering published abundances, it is best to use Ba abundances based on weak lines or to substitute the abundance of the s-process element La in place of Ba, because lanthanum is dominated by a single isotope (99.9% ¹³⁹La) with relatively weak lines.

Many astrophysical environments have been proposed as the main source of the r-process. Mathews & Cowan (1990), Mathews et al (1992) list many of these and attempted to test the possibilities by comparing predicted abundances from a Simple model of Galactic chemical evolution to observed heavy element abundances. They claimed that low-mass SN II (7-8 M) were the most likely candidates. In their model the trend of increasing [heavy element/Fe] ratios with [Fe/H] was due to a time delay arising from the longer main-sequence lifetime of low-mass SN II progenitors relative to high-mass SN II progenitors. Thus at early times, when the lowest metallicity prevailed, only high mass SN II occurred with low [heavy element/Fe] yield ratios; at later times, and higher metallicity, the low mass SN II enriched the Galactic gas with high [heavy element/Fe] material. This model requires that in all situations the first low-mass SN II events were preceeded by high-mass SN II events, which probably would not occur in the case of chemical evolution in molecular cloud size masses or Searle-Zinn fragments (Searle & Zinn 1978). The model also requires that the mean [Sr/Fe] ratio increases with increasing [Fe/H] at low metallicity; however, the results of McWilliam et al (1995b) indicate a constant average [Sr/Fe] value. Thus, the time-delay mechanism cannot be used to explain the observed heavy element abundances in the halo, and no constraint can be placed on the mass of the SN chiefly responsible for heavy element synthesis.

McWilliam et al (1995a, b, 1996), Sneden et al (1994) argued that the observed dispersion in heavy element abundances must reflect an intrinsic dispersion in the [heavy element/Fe] ratio of the gas from which the extremely metal-poor halo stars formed. In particular, the r-process abundance pattern and the high frequency of stars with heavy element enhancements rules out the possibility that these stars are population II barium stars.

The heavy element dispersions found by McWilliam et al (1995a, b) showed that the range in SN heavy element yields is at least a factor of 300. McWilliam et al (1996) argued that because the heavy element/Fe ratio for CS 22892-052 is ~ 15 times the asymptotic value, the progenitor SN must represent no more than 1/15 of all SN II. Because homogenization of the halo gas could only have occurred once the full range of SN yields was sampled, the metallicity of the homogenization point (at [Fe/H] = -2.5) corresponds to approximately 15 SN events. If 0.1 M of iron is ejected per SN II event, then this metallicity requires mixing of the ejecta with ~ 10⁵ to 10⁶ M of hydrogen.

Searle & McWilliam (1997 in progress) have studied models of chemical enrichment by small numbers of SN II events, with [Sr/Fe] yields selected at random from the observed range in [Sr/Fe]. This stochastic model can reproduce the average, the dispersion, and the envelope of [Sr/Fe] values seen in metal-poor halo stars, with [Fe/H] -2.5. The model is consistent with enrichment by single SN II events below [Fe/H] ~ -3.3, in regions of mass ~ 10⁶ M, which is characteristic of present-day molecular clouds.

The large r-process enhancements in CS 22892-052 allowed Sneden et al (1996) to measure the abundance of thorium (Th, a pure r-process element) in this star. Owing to its 14 × 10⁹ year half life, Th is potentially a useful Galactic chronometer (e.g. Butcher 1987). Based on the solar [Th/Eu] ratio, Sneden et al (1996) deduced a minimum age for CS 22892-052 of 15 ± 4 × 10⁹ years. Cowan et al (1997) employed r-process nucleosynthesis calculations and various Galactic chemical evolution models to predict the initial r-process [Th/Eu] ratio, which led to a minimum age of 15 ± 4 × 10⁹ years and a most likely age of 17 ± 4 × 10⁹ years.

The early work of Pilachowski et al (1983) indicated low, and even subsolar, [Eu/Fe] ratios for globular cluster stars; taken at face value these results suggest a difference between the composition of halo field stars and globular cluster stars. However, the more recent of François (1991), Brown & Wallerstein (1992), Shetrone (1996a), McWilliam et al (1992) all indicate globular cluster [Eu/Fe] ratios near +0.4 dex, which is similar to the results for halo field stars (e.g. Gratton & Sneden 1994, Shetrone 1996a, McWilliam et al 1995a, b, Magain 1989, recomputed here); the mean halo [Eu/Fe] value is +0.33 dex, which is the same enhancement as seen in the most metal-poor disk stars (Woolf et al 1995).

The chemical composition of the unusual globular cluster Cen differs from other globular clusters and field halo stars. Several recent abundance studies of Cen giants have been published: Vanture et al (1994), Norris & Da Costa (1995), Smith et al (1995), Norris et al (1996). The cluster shows a metallicity spread from [Fe/H] = -1.9 to -0.6, with evidence for two star formation epochs. The [ / Fe] ratios show the normal factor of 2 enhancement seen in halo stars, which implicates nucleosynthesis by SN II only. However, the heavy elements are enhanced well above the solar value and are consistent with significant contamination by s-process nucleosynthesis from AGB stars. This is evidence that the s-process occurs more rapidly than the time scale for enrichment by SN Ia. A puzzle noted by Smith et al (1995) is the subsolar [Eu/Fe] ratio, near -0.4 dex; if AGB stars produced s-process material, then SN II should have produced larger [Eu/Fe] ratios. It is as if the r-process SN II never occurred in Cen; perhaps this is an indication of a unique IMF that excluded r-process SN II events.

² Barium stars are thought to arise from mass transfer from an AGB star that has polluted its envelope with s-process elements (e.g. McClure 1984). Back.