Lithium-7 is fragile, burning in stars at a relatively low temperature. As a result, the majority of any interstellar 7Li cycled through stars is destroyed. For the same reason, it is difficult for stars to create new 7Li and/or to return any newly synthesized 7Li to the ISM before it is destroyed by nuclear burning. In addition to synthesis in stars, the intermediate-mass nuclides 6Li, 7Li, 9Be, 10B, and 11B can be synthesized via cosmic ray nucleosynthesis, either by alpha-alpha fusion reactions, or by spallation reactions (nuclear breakup) in collisions between protons and alpha particles and CNO nuclei. In the early Galaxy, when the metallicity is low, the post-BBN production of lithium is expected to be subdominant to that from BBN abundance. As the data in Figure 9 reveal, only relatively late in the evolution of the Galaxy does the lithium abundance increase. The data also confirm the anticipated "Spite plateau" (Spite & Spite 1982), the absence of a significant slope in the Li/H versus [Fe/H] relation at low metallicity due to the dominance of BBN-produced 7Li. The plateau is a clear signal of the primordial lithium abundance. Notice, also, the enormous spread among the lithium abundances at higher metallicity. This range in Li/H likely results from the destruction/dilution of lithium on the surfaces of the observed stars while they are on the main sequence and/or lithium destruction during their pre-main sequence evolution, implying that it is the upper envelope of the Li/H versus [Fe/H] relation that preserves the history of Galactic lithium evolution. Note, also, that at low metallicity the dispersion is much narrower, suggesting that corrections for depletion/dilution are (may be) much smaller for the Population II stars.
Figure 9. A compilation of the lithium abundance data as a function of metallicity from stellar observations (courtesy of V. V. Smith). (Li) 1012(Li/H), and [Fe/H] is the usual logarithmic metallicity relative to solar. Note the "Spite plateau" in Li/H for [Fe/H] -2.
As with the other relic nuclides, the dominant uncertainties in estimating the primordial abundance of 7Li are not statistical, but systematic. The lithium observed in the atmospheres of cool, metal-poor, Population II halo stars is most relevant for determining the BBN 7Li abundance. Uncertainties in the lithium equivalent width measurements, in the temperature scales for the cool Population II stars, and in their model atmospheres dominate the overall error budget. For example, Ryan et al. (2000), using the Ryan, Norris, & Beers (1999) data, infer [Li]P 12 + log(Li/H) = 2.1, while Bonifacio & Molaro (1997) and Bonifacio, Molaro, & Pasquini (1997) derive [Li]P = 2.2, and Thorburn (1994) finds [Li]P = 2.3. From recent observations of stars in a metal-poor globular cluster, Bonifacio et al. (2002) derive [Li]P = 2.34 ± 0.056. As may be seen from Figure 9, the indication from the preliminary data assembled by V. V. Smith (private communication) favors a Spite plateau at [Li]P 2.2.
In addition to these intrinsic uncertainties, there are others associated with stellar structure and evolution. The metal-poor halo stars that define the primordial lithium plateau are very old. As a result, they have had time to disturb the prestellar lithium that could survive in their cooler, outer layers. Mixing of these outer layers with the hotter interior where lithium has been (can be) destroyed will dilute or deplete the surface lithium abundance. Pinsonneault et al. (1999, 2002) have shown that rotational mixing may decrease the surface abundance of lithium in these Population II stars by 0.1 - 0.3 dex while still maintaining the rather narrow dispersion among the plateau abundances (see also Chaboyer et al. 1992; Theado & Vauclair 2001; Salaris & Weiss 2002). Pinsonneault et al. (2002) adopted for a baseline (Spite plateau) estimate [Li] = 2.2 ± 0.1, while for an overall depletion factor 0.2 ± 0.1 dex was chosen. Adding these contributions to the log of the primordial lithium abundance linearly, an estimate [Li]P = 2.4 ± 0.2 was derived. In the comparison between theory and observation below, I will adopt the Ryan et al. (2000) estimate [Li]P = 2.1 ± 0.1, but I will also consider the implications of the Pinsonneault et al. (2002) value.