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

9. IRON-PEAK ELEMENTS

The iron-peak elemental abundances showing conclusive evidence for deviations from the solar ratios are summarized in Figure 12. Wallerstein (1962), Wallerstein et al (1963) were the first to find evidence of a nonsolar mixture of iron-peak elements: Deficiencies of Mn found by Wallerstein were confirmed by later studies (e.g. Gratton 1989). From [Fe/H] = 0.0 to -1.0, the [Mn/Fe] ratios are deficient in a manner that mirrors the -element overabundances, and in the interval [Fe/H] = -1.0 to -2.5 dex, [Mn/Fe] is constant at ~ -0.35 dex. Thus the [Mn/Fe] trend is similar, but in an opposite sense to the [ / Fe] trend with [Fe/H]. A simple conclusion is that a significant source of Mn comes from SN Ia. McWilliam et al (1995a) discovered that below [Fe/H] ~ -2.5, the [Mn/Fe] ratio decreases steadily with decreasing [Fe/H], like the trends exhibited by the heavy elements.

Figure 12. Iron-peak abundance trends for elements that deviate from the solar [M/Fe] ratios. (a-c) Halo field stars: filled circles (McWilliam et al 1995), open squares (Gratton & Sneden 1988, 1991, Gratton 1988). Note that Cr, Co, and Mn each show a decline relative to Fe below [Fe/H] = -2.5; [Mn/Fe] also declines between [Fe/H] = 0.0 to -1. (d) [Cu/Fe] versus [Fe/H] from Sneden et al (1991): filled circles (field stars), filled triangles (globular clusters), open triangles (reanalyzed field star data), open circles (literature values for population I field stars).

Wallerstein (1962, 1963) observation of Mn deficiencies at low metallicity was claimed to be part of the neutron-excess-dependent yields of Arnett (1971). Even the recent Galactic nucleosynthesis predictions of Timmes et al (1995), based on the Woosley & Weaver (1995) calculations for SN II, predict deficiencies of Sc, V, Mn, and Co of ~ -0.5 dex relative to Fe for metal-poor stars. Wallerstein claimed that V is also deficient in metal-poor stars, but this finding was not confirmed by subsequent analyses (e.g. Pagel 1968). The abundance of V has not been well studied. The most comprehensive analysis was done by Gratton & Sneden (1991), who found [V/Fe] ~ 0.0 at all metallicities, which confirmed Pagel's conclusion. If the Mn deficiencies are due to a neutron-excess dependence, then V and Sc are also expected to follow the same trend, which is not observed.

Sneden & Crocker (1988), Sneden et al (1991a) studied the abundances of Cu and Zn as a function of metallicity and discovered that [Cu/Fe] decreases linearly with declining metallicity, [Cu/Fe] = 0.38[Fe/H] + 0.15 (curiously the trend resembles that of [Al/Fe] with [Fe/H]), while Zn is constant, at [Zn/Fe] = 0.0, for all metallicities. Sneden et al suggested that nucleosynthesis of Cu may occur mainly by the weak s-process in the cores of massive stars, with a small contribution from explosive burning in SN II. However, Matteucci et al (1993) suggested that the greatest production of Cu and Zn occurs in SN Ia. If this is true, then some SN Ia occurred for [Fe/H] < -1, which will have important consequences for chemical evolution models of the halo.

The constant [Zn/Fe] abundance ratio seen in the Galaxy is not universal: Abundance analyses of QSO absorption line systems show that Zn is enhanced relative to Fe (e.g. Pettini et al 1994, Lu et al 1996). Although the observed enhancements of the [Zn/Cr] ratios may indicate that the gas has been affected by dust depletion, if this were the case, then one would also expect to find large [S/Cr] ratios, which are not found. Thus the enhanced Zn abundance in QSO absorption line systems may have a nucleosynthesis origin.

Besides Zn, the abundance ratios of Sc, V, and Ni relative to Fe seem to scale with [Fe/H]. It should be noted that Zhao & Magain (1989, 1990) claimed a mean +0.27-dex enhancement of [Sc/Fe] in metal-poor dwarfs. However, high quality data of Gratton & Sneden (1991), Peterson et al (1990), as well as the results of McWilliam et al (1995b) found no evidence for a deviation from [Sc/Fe] = 0.0 in metal-poor giant stars. The lower quality data of Gilroy et al (1988) actually indicated a deficiency of ~ 0.2 dex.

Luck & Bond (1985) claimed enhanced [Ni/Fe] ratios in metal-poor stars, and Pilachowski et al (1996) found a mean [Ni/Fe] = -0.27 near [Fe/H] = -2; however, other studies tended to find solar [Ni/Fe] ratios everywhere. In particular, Peterson et al (1990) demonstrated that the Luck & Bond Ni overabundances were probably the result of selecting lines enhanced above the detection threshold by noise-spikes. The combined studies of Gratton & Sneden (1991), Peterson et al (1990), Edvardsson et al (1993), McWilliam et al (1995a), Ryan et al (1996) are inconsistent with [Ni/Fe] more than ± 0.1 dex from the solar ratio in the interval -4 [Fe/H] 0; although some of the Gratton & Sneden points are subsolar near [Fe/H] = -2.5.

Until recently, [Co/Fe] and [Cr/Fe] ratios were commonly accepted to be independent of [Fe/H]; however, McWilliam et al (1995b) showed that Co and Cr deviate from a plateau at metallicities below [Fe/H] ~ -2.5 (see Figure 12). McWilliam et al (1995a) found evidence supporting their results in the data of Gratton & Sneden (1991), Ryan et al (1991), and Wallerstein et al (1963); these trends for Co, Cr, and Mn have subsequently been verified by Ryan et al (1996).

The divergence of [Co/Fe], [Mn/Fe], and [Cr/Fe] (and the heavy elements) from a plateau, below [Fe/H] ~ -2.5, suggests that chemical evolution was very different below the lowest globular cluster metallicities, perhaps indicating the existence of population III or early population II stars. If SN II were the dominant source of iron-peak elements at low metallicity, the observed range in [M/Fe] ratios indicates the minimum range of yield ratios for SN II.

McWilliam et al (1995b, 1996) argued that because the [Co/Fe], [Mn/Fe], [Cr/Fe], and heavy element/Fe ratios differ from the solar values, the metal-poor stars below [Fe/H] ~ -2.5 cannot be the products of simple dilution of higher metallicity gas (e.g. typical of globular cluster composition) by gas with zero metallicity. If metal-poor stars below -2.5 were products of dilution with pure hydrogen, then solar [Co/Cr] ratios would exist down to -4. In Figure 13, the tight correlation between [Co/Cr] and [Fe/H] sets tight constraints on the dispersion in dilution by zero metallicity gas that might have occurred, i.e. dilution with zero metal gas could not have differed from one star forming region to another by more than a factor of 2 because the [Fe/H] dispersion at fixed [Co/Cr] value is ~ 0.3 dex.

Figure 13. [Co/Cr] versus [Fe/H] for extremely metal-poor halo stars: crosses from McWilliam et al (1995), open boxes from Ryan et al (1991). The lines trace the resultant [Co/Cr] ratios when solar-composition material is mixed with primordial material, characterized by a large [Co/Cr] value near +1.3 dex.

It would be interesting to understand why the [Co/Cr] ratios of extremely metal-poor stars are so tightly correlated with [Fe/H], whereas for the same stars, the heavy elements show such a large dispersion (Figure 9).

McWilliam et al (1995b) suggested that the heavy elements are produced in large amounts by a rare subclass of SN event and essentially not at all in most events. The tight correlation of [Co/Cr] with [Fe/H] might suggest that all or most SN produce iron-peak elements, like Cr, Mn, and Co, in similar proportions but with metallicity-dependent yields. One possibility suggested by McWilliam et al (1995b) is that the observed [Co/Cr] trend in Figure 13 could have occurred if the star-forming gas was in a process of steady chemical enrichment, with several generations of SN gradually enriching the parent cloud in the range -4 [Fe/H] -2.5. In this model, different kinds of SN II existed with different [Co/Cr] yield ratios, and as metallicity increased, the average SN II changed in character until, when metallicity reached [Fe/H] ~ -2.5, the average SN produced solar [Co/Cr] yield ratios.

One way for the metallicity-modulated [Co/Cr] yields to occur may be by affecting the mean SN progenitor mass. For example, Wolfire & Cassinelli (1987), Yoshii & Saio (1986), Kahn (1974) predicted that the IMF is weighted to higher mass stars at low metallicity. Another potential method of altering the SN progenitor mass range is through the effect of metallicity-dependent mass loss; for example, theoretical models by Bowen & Willson (1991) suggested that AGB stars with an initial mass of 2.4 M reached the Chandrasekhar mass before the envelope could be ejected.

The observed [Co/Cr] versus [Fe/H] relation in Figure 13 cannot be explained by the time delay between SN II progenitors with different mass, for the following reason: Any time delay must be less than the maximim SN II progenitor lifetime, which, at 20 × 10⁶ years, is much shorter than the dynamical time scale of the Galaxy; thus chemical evolution in the first 20 × 10⁶ years must have taken place in isolated regions. The observed range of [Fe/H] at a fixed [Co/Cr] in Figure 13 excludes a range of SFR more than a factor of 2, which conflicts with the fact that SFR are known to vary by orders of magnitude.

Ryan et al (1996) suggested that the SN II [Co/Cr] yield is a function of the SN energy and that the SN energy dictates how much dilution of the ejecta occurs and, therefore, the metallicity of the next generation of stars. In this way the observed tight correlation between [Co/Cr] and [Fe/H] can occur.

An alternative mechanism (L Searle & A McWilliam, in preparation) is that there was a primordial composition, characterized by high [Co/Cr], and that this was later diluted with solar composition SN ejecta. In this model, the first generation SN, presumably zero-metallicity population III stars, produced high [Co/Cr] ratios, but all subsequent SN II produced solar [Co/Cr] yield ratios. Figure 13, shows two "dilution" curves that follow the combined composition of primordial plus solar mix material, with increasing amounts of solar composition material. It is clear that both dilution curves fit some but not all of the observed data points, which may suggest that the putative primordial material was characterized by a spread in metallicity as well as a high [Co/Cr] ratio, near +1 dex.

This mechanism produces a tight correlation of [Co/Cr] with [Fe/H], even with single SN events; thus it fits into the model of discrete chemical enrichment used by Searle & McWilliam to explain the observed heavy element dispersion. This model implies that stars with [Fe/H] -3.3 may be the products of individual SN events.

The mechanisms proposed by Ryan et al (1996) and Searle & McWilliam are roughly consistent with the predictions of Audouze & Silk (1995), who considered the physics of mixing of SN ejecta and concluded that there must be a minimum possible metallicity, near [Fe/H] = -4.

An unusual kind of variation in iron-peak elements was found for the star CS22949-037 by McWilliam et al (1995b): The elements outside the range from Ti to Ni (i.e. C, Na, Mg, Al, Si, Ca, Sc, Sr, Ba) appear to be overabundant by typically ~ 1 dex. The elements within the range Ti to Ni exhibit normal relative ratios. McWilliam et al (1995b) suggest that a simple explanation is that this star is actually deficient in the iron-peak elements. This observation suggests that SN exist that produce relatively small iron-peak yields. Ryan et al (1996) have found two more stars, CS 22876-032 and CS 22897-008, that show unusual chemical compositions, which are indicative of star-to-star scatter and a dispersion in element yields. These star-to-star variations indicate that at low metallicity, the intrinsic dispersion in SN yields can produce anomalous stellar compositions because the averaging process of combining yields from many SN is not yet complete.