Recent studies2,3,4 have suggested that the transport processes accompanying a deep mining phase, called the third dredge-up, could induce some partial mixing of protons below the convective hydrogen envelope into the helium-burning carbon-rich layers. This partial mixing of protons triggers the chain of reactions 12C(p,γ)13N(β)13C(α,n)16O that is held responsible for the large neutron irradiation required to synthesize heavy elements by the s-process.

Although this scenario is nowadays the most widely accepted one, its main ingredient (the partial mixing of protons in the deep carbon-rich layers) cannot yet be derived from first physical principles. It is therefore of prime importance to devise predictions that may test this proton-mixing scenario against abundance observations. One such prediction is that low-metallicity AGB stars should exhibit large overabundances of Pb–Bi as compared to lighter s-elements. Such stars would be called ‘Pb stars’1. They are characterized not only by large [Pb/Fe] and [s/Fe] abundance ratios, but also by large [Pb/s] abundance ratios, where [A/X] ≡ log(NA/NX) - log(NA/NX), s stands for any element produced by the s-process and subscript refers to the Solar System abundances. In particular, [Pb/hs] (where hs denotes the ‘heavy’ s-process elements such as Ba, La or Ce) ratios as large as 1.5 are predicted1 in AGB stars with [Fe/H] -1.3. Within the framework of the partial mixing of protons into the deep carbon-rich layers, this particular prediction is found to be quite robust1 with respect to the model parameters (like the abundance profile of the protons in the partially mixed layers, or the extent of the partial mixing zone) and uncertainties (for example, reaction rates). All s-process-enriched AGB stars with metallicities [Fe/H] -1.3 are thus predicted to be Pb stars, independently of their mass or metallicity (provided the partial mixing of protons takes place).

However, Pb stars remained elusive: none of the low-metallicity stars with large Pb overabundances previously reported in the literature5,6,7 met all the above criteria. Indeed, the Pb abundances5,6 in the metal-poor stars HD 115444, HD 126238 and CS 22892-052 are attributed almost exclusively to the r-process, rather than the s-process, because they occur with strong enhancements of r-elements such as Eu (the r-process is an alternative neutron-capture process occurring in high-neutron flux environments like supernova explosions). In the very metal-poor ([Fe/H] = -2.7) carbon- and s-process-rich star LP 625-44, lead was found7 to be largely overabundant with respect to iron ([Pb/Fe] = +2.65), but not with respect to other s-elements (for example, [Pb/La] = 0.15, [Pb/Ba] = -0.09), in disagreement with the model predictions.

In the course of a programme of high-resolution spectroscopic observations at the European Southern Observatory 3.6-m telescope, we have discovered three stars that meet all the conditions to be classified as Pb stars.

This observing programme focuses on the study of the nucleosynthesis of elements heavier than iron—with special emphasis on lead—in s-process-rich, low-metallicity AGB stars. It ultimately aims at identifying the origin of Pb–Bi, the heaviest stable elements in the Universe, and at evaluating the respective contributions of the s- and r-processes to their galactic content. This information provides interesting constraints on the operation of these nucleosynthesis processes in stars, and in particular constitutes an important ingredient in the prediction of stellar ages based on the actinide cosmochronometry8.

As low-metallicity AGB stars seem to be very rare in the solar neighbourhood nowadays, the family of CH stars represents a good alternative for the purpose of studying the signature of low-metallicity s-processing. These low-metallicity stars are members of binary systems9, and their heavy-element overabundances are caused by contamination from ejecta of their companion, formerly an AGB star, presently a dim white dwarf. The surface composition of CH stars thus traces the nucleosynthesis that occurred in a given low-metallicity AGB star. Comparison with model predictions is therefore straightforward, which would not be the case if we were to use surface abundances that would be derived from low-metallicity unevolved field stars (their abundances reflect, on the contrary, the plethora of nucleosynthetic events that shaped the composition of the interstellar medium from which the stars formed).

All three stars observed present large Pb abundances (Fig. 1 and Table 1) and may be considered to be genuine s-process Pb stars, because (1) these s-process-rich stars do exhibit overabundances of Pb with respect to all the other s-process elements, and (2) the Pb overabundance cannot originate from the competing r-process (that hypothesis would indeed lead to an abundance pattern totally incompatible with the observed one; see Fig. 2). The estimated errors on our abundance determinations do not endanger this conclusion (Tables 1 and 2). High spectral resolution is essential to resolve the 405.781-nm Pb line from the nearby CH line at 405.77 nm, and therefore to derive correct lead abundances (see Fig. 1).

Figure 1: Comparison between observed and synthetic spectra of HD 196944 around the Pb I line at 405.781 nm.
figure 1

The observed spectrum (dots) has been obtained at the ESO 3.6-m telescope and Coudé Echelle Spectrometer delivering a spectral resolution λλ of 135,000 at 405 nm. Synthetic spectra corresponding to [s/Fe] = 0 (thin solid line; in this case the Pb line is not visible) and [s/Fe] given in Table 1 (thick solid line) are also shown.

Table 1 Derived element abundances in three CH stars
Figure 2: Comparison of the observed abundances with predictions obtained in the framework of the proton-mixing scenario.
figure 2

The abundances observed in HD196944 (panel a) and in HD187861 and HD224959 (panel b) are compared with the model abundances obtained as described in ref. 1 for stars of the same metallicities as the programme stars. The abundance scale log(εX) stands for log(NX/NH) + 12. The error bars on the observations are the root mean square of the uncertainties due (1) to the spectral fit and to the dispersion on the derived abundances when several lines of a given element were used and (2) to the adopted model parameters, that is, the effective temperature, gravity and microturbulence (see Table 2). Panel a also shows the solar system r-abundance distribution normalized to Ce. This comparison clearly shows that the large [Pb/hs] ≈ 1.2 abundance ratio cannot be understood in terms of an r-process enrichment, of which the solar r-process abundances may be considered as a fairly typical representation5.

Table 2 Effect of uncertainties in model atmosphere parameters on the element abundances derived for HD196944

The observed surface abundances are in remarkable agreement with the predictions1 obtained for partial mixing operating in stars of the corresponding metallicities (Fig. 2). As indicated above, there is actually little room for variations at these low metallicities within the framework of the ‘proton-mixing’ scenario (in that sense, the LP 625-44 abundance pattern7 appears puzzling, and may call for an alternative scenario10 operating in extremely low-metallicity stars, provided LP 625-44 is a representative case).

The similarity between the abundance patterns of our three stars is remarkable, and strongly suggests that this pattern constitutes the standard signature of s-processing at low metallicity, in agreement with the model predictions described above. If so, the discovery of the three Pb stars reported here validates the ‘proton-mixing’ scenario for the detailed operation of the s-process in AGB stars. This model is currently the only one capable of explaining such high Pb overabundances in s-process-enriched stars (see Fig. 2). As a consequence, the solar s-process distribution should no longer be thought to result from successive neutron irradiations within a single star11, but rather as a result of the chemical evolution of the Galaxy shaping the s-process abundance distribution12.

Moreover, to calculate stellar ages cosmochronologically on the basis of actinide abundances we need to separate the contributions of the s- and r-processes to the galactic Pb production. The discovery of low-metallicity s-rich Pb stars makes it possible to constrain the r-process contribution, which controls our prediction of the production of the long-lived actinides. More accurate ages for the oldest stars in our Galaxy using the Th cosmochronometry may thus be obtained8.