Annu. Rev. Astron. Astrophys. 1981. 19: 77-113 Copyright © 1981 by Annual Reviews. All rights reserved

2.4. Primordial Variations in Relative Abundances

Primordial variations in relative abundances of different nuclear species are expected as a consequence of the existence of various nucleosynthesis processes taking place in different sites and, therefore, affecting the interstellar medium at different rates. However, such variations (as opposed to those caused by self enrichment affecting stellar atmospheres in later phases of evolution) have proved quite difficult to establish and for a long time it was possible to say that all the elements from carbon upwards vary in lockstep, within the uncertainties of observation (e.g. Unsöld 1974), and also that for moderate values of [Fe/H], such as found in disk stars, [Fe/H] can still be regarded as a good indication of relative heavy-element abundance although carbon and oxygen are the main individual contributors to this (about 20 and 45 percent, respectively). In the more extreme cases, however, primordial variations in some elements and even isotopes are now well established (cf. Pagel 1979a). In what follows, we briefly summarize the present status of primordial abundance variations in individual elements relative to Fe or other standards when appropriate.

Evolution of the light elements (Li, Be, B), which are affected by destruction at stellar surfaces, has been discussed by Reeves & Meyer (1978). Carbon shows little or no specific abundance variation in unevolved field stars (Peterson & Sneden 1978 and references therein) whereas nitrogen shows a significant tendency to be deficient in weak-metal stars, though with exceptions (Harmer & Pagel 1970, 1973, Tomkin & Bell 1973, Sneden 1974, Clegg 1977, Sneden & Peterson 1977, Pagel 1979a). Variations in N/O in the interstellar medium are discussed below. Oxygen in weak-metal stars is usually deficient by smaller factors than iron, with [O/Fe] 0.6 (Sneden, Lambert & Whitaker 1979), and observations of halo planetary nebulae suggest a similar result for both O and Ne (Peimbert 1978, Kaler 1980). These generalizations do not necessarily apply to the stars in globular clusters, where some of the variations in CH, CN, and CO seem to be too drastic - and to extend too far down towards the main sequence - to be explained by mixing effects alone (see Kraft 1979; but cf. Cohen 1980a) and positive [O/Fe] values have not been found in M13 (Pilachowski, Wallerstein & Leep 1980), although they are found, or may be present, in several other clusters such as M92, M15 (Cohen 1979b), M3, M5 (Pilachowski, Wallerstein & Leep 1980), NGC 288 (Pilachowski & Sneden 1980), M71, and Cen (Cohen 1980a, b, Mallia & Pagel 1981). The primordial (C + N + O)/Fe ratio is probably not the "second parameter" affecting horizontal-branch types (cf. Kraft 1979, Pilachowski, Wallerstein & Leep 1980, Zinn 1980b) because both M13 and NGC 288 are of the "blue" HB type (Pilachowski & Sneden 1980).

Elements from sodium to silicon show odd-even effects in some of the most metal-deficient stars, as is predicted by the theory of explosive carbon burning (Arnett 1971; but cf. Arnett & Wefel 1978), superposed on a small enhancement of the "-particle" elements Mg, Si, Ca, Ti relative to iron (Wallerstein 1962, Peterson 1981 and references therein). Particularly striking is the factor of 4 deficiency of ²⁵Mg and ²⁶Mg relative to ²⁴Mg in the subdwarf Groombridge 1830 (Tomkin & Lambert 1980). When many weak-metal stars are considered together there is a large scatter in Na/Fe and especially Al/Fe (Peterson 1981; cf. also Spite & Spite 1978, 1980). Spite & Spite have suggested that Al/Fe may be influenced by diffusion effects, and this receives some support from the fact that Peterson's Al-deficient stars are confined to a narrow range of effective temperatures, between 5800 and 6300 K. In NGC 6752, Al I features are positively correlated with CN (Norris et al. 1980), while in M13 three stars show large Na/Fe variations which may be correlated with carbon or anti-correlated with oxygen (or both) (Peterson 1980) although both clusters are quite homogeneous in [Fe/H].

The behavior of calcium is of some importance because of the use of the spectral-type difference S as an abundance indicator in field and cluster RR Lyrae stars. Various analyses lead to an apparent spread in [Ca/Fe] at given [Fe/H] (Butler, Dickens & Epps 1978), which could be largely spurious (Butler & Deming 1979, Suntzeff 1980); Ca/Fe is then enhanced by factors close to 2 for metal-deficient stars in general, as shown in the diagram by Peterson (1981). The same applies to titanium, Variations within the iron group are virtually insignificant (Pagel 1979a), a conclusion now extended to copper (Peterson 1981).

Elements which (in the Solar System at least) are predominantly due to the s-process show a barely significant correlation with [Fe/H] for disk stars (Spite 1968, Huggins & Williams 1974) and a more significant correlation for extreme halo stars (Spite & Spite 1978); yttrium, which may be taken as typical of the N = 50 magic number peak, varies with about half of the amplitude of barium, on the N = 82 peak, and the latter roughly follows the relation

[Ba/Fe] 0;	[Fe/H] > -1.3
[Ba/Fe] = 0.6([Fe/H] + 1.3);	[Fe/H] -1.3.

In stating these conclusions, we are assuming that all of the so-called barium stars, with [Ba/Fe] > 0, are self-enriched, which may not be completely true in the light of the recent result that extreme (but not mild) Ba II stars are binary (McClure, Fletcher & Nemec 1980). Europium, which is the only readily accessible r-process element in stellar spectra, varies in lockstep with iron quite closely (Butcher 1972, Spite & Spite 1978). The variations in s- and r-process elements, therefore, are qualitatively as expected if the latter are primary nucleosynthesis products from supernovae which also make iron and the former are secondary products of red giant evolution with mixing (typified by carbon, S, and Ba II stars), but quantitatively the s-process variations are far smaller than would be expected in a situation of steady enrichment. Models of galactic chemical evolution have yet to come to grips with this problem (cf. Tinsley 1979, 1980).