Main

We obtained high-resolution spectra (R ≈ 70,000) of the star Sk 143 (AzV 456), an O-type supergiant star in the Small Magellanic Cloud (SMC), using the Ultraviolet and Visual Echelle Spectrograph (UVES)12 on the 8.2-m Very Large Telescope (VLT); observational details are given in the Supplementary Information. The sight line to this star was chosen for observation because it shows significant absorption from neutral atoms and molecules13,14,15 and a weak interstellar radiation field14, all of which favour the presence of neutral lithium (Li i). Li i absorption is clearly detected along this sight line (Fig. 1).

Figure 1: Interstellar absorption by several neutral species seen towards the star Sk 143.
figure 1

Normalized interstellar absorption profiles from UVES plotted versus the Local Standard of Rest velocity, vLSR, and profile fit of the Li i absorption. The empirically determined signal-to-noise ratio is about 275 per pixel (5 pixels per resolution element) for the Li i observations. The full set of optical and ultraviolet absorption profiles seen towards this star and the column densities measured from these are given in the Supplementary Information. b, The profiles of Li i, K i, and Fe i; the SMC cloud bearing Li i at vLSR ≈ +121 km s−1 is marked with the dashed line. The thicker grey regions near Li i are possibly contaminated by diffuse interstellar bands or residual fringing, which may extend into the region containing Li absorption. The effects on the 7Li i columns are within the quoted uncertainties. The Li i absorption is composed of (hyper)fine structure components of both 7Li i and 6Li i (shown, respectively, by the green and blue ticks in the top panel of a). The strong line of 7Li i is detected with approximately 16σ significance in the ISM of the SMC. A model fit to the Li i absorption complex is shown in a (see Supplementary Information), with the difference between the data and the fit, δ, shown immediately below (normalized to the local error array). The free parameters for the fit are the polynomial coefficients for the stellar continuum, the central velocity, Doppler parameter (b-value), and column densities of 7Li i and 6Li i for the interstellar cloud. The red curve shows the best-fitting model including both 7Li i and 6Li i, which are shown in green and blue, respectively. The best-fit isotopic ratio is N(6Li i)/N(7Li i) = 0.13 ± 0.05 (68% confidence limit), consistent with the presence of 6Li along the sight line, although below the 3σ detection threshold.

PowerPoint slide

Full size image

The derivation of the total Li/H abundance in the interstellar medium (ISM) requires large corrections for ionization, given the column density of Li, N(Li) ≈ N(Li ii) ≫ N(Li i), and for the incorporation of Li into interstellar dust grains16. Our first approach to these corrections uses observations of adjacent ionization states of other metals, in this case Ca and Fe, to estimate the amount of unseen gas-phase lithium. Assuming ionization balance and only atomic processes, we have the ratios N(Li ii)/N(Li i) ∝ N(Ca ii)/N(Ca i) or N(Li ii)/N(Li i) ∝ N(Fe ii)/N(Fe i), where the constants of proportionality involve the ratios of ionization rates and recombination coefficients for the elements in question16,17. The ratio of 7Li i to total hydrogen in the SMC is log[N(7Li i)/N(H)] = –11.17 ± 0.04 (all uncertainties are 1σ unless noted), where N(H) ≡ N(H i) + 2N(H2). Applying ionization corrections derived from Ca and Fe yields logarithmic abundances A(7Li) ≡ log[N(7Li)/N(H)] + 12 = 2.79 ± 0.11 (from Ca) and 3.01 ± 0.12 (from Fe). These calculations do not include more complicated (and uncertain) effects such as grain-assisted recombination17,18, nor do they correct for dust depletion.

Our second approach uses the observation16 that N(7Li i)/N(K i) along sight lines through the Milky Way is nearly constant (with new determinations giving consistent results19,20). When a differential ionization correction is applied, 7Li/K in the Milky Way ISM is consistent with the Solar System ratio. Thus, 7Li and K appear to have very similar ionization and dust depletion behaviours, and 7Li i/K i gives a good measure of the total (gas+dust phase) 7Li/K (refs 16, 19 and 20). We measure log[N(7Li i)/N(K i)] = −2.27 ± 0.03 in the SMC, in agreement with the Galactic relationship19,20. Applying an ionization correction of +0.54 ± 0.08 dex (refs 16 and 17) gives log[N(7Li)/N(K)] = −1.78 ± 0.09. With the Solar System ratio derived from meteorites21, we find . The ratio of 7Li to metal nuclei in the SMC is consistent with that found in the Solar System and the Milky Way ISM16: .

Although the ionization and depletion characteristics of S i are not as well tied to those of Li i (ref. 17), a similar approach using S i yields [7Li/S]SMC = −0.26 ± 0.11. The sub-solar ratio is consistent with a modest (0.3 dex) depletion of Li and K onto dust in the ISM19 relative to S.

We estimate A(7Li) by scaling 7Li/K to Li/H: . We adopt [7Li/K]SMC from above, the meteoritic (ref. 21), with a mean present-day SMC metallicity [Fe/H]SMC = −0.59 ± 0.06 and an SMC K/Fe abundance [K/Fe]SMC ≡ +0.00 ± 0.10 (these last two are discussed in the Supplementary Information). This yields A(7Li)SMC = 2.68 ± 0.16. Similarly scaling the 7Li/S result gives 2.38 ± 0.17.

Most previous observational constraints on the primordial Li abundance have relied on measurements of atmospheric abundances in low-metallicity Galactic stars. Our detection of interstellar lithium beyond the Milky Way opens a new window on the lithium problem. Although there are significant uncertainties associated with ionization and dust effects, as demonstrated by the significant spread in A(7Li)SMC values, these are largely independent of the uncertainties that might affect stellar measurements of the primordial lithium abundance. Our recommended absolute abundance is A(7Li)SMC = 2.68 ± 0.16, or (7Li/H)SMC = (4.8 ± 1.8) × 10−10, derived from 7Li/K. This is compared to stellar 7Li abundances6,22 at different metallicities in Fig. 2. Our best estimate overlaps the prediction from standard BBN using the baryonic density deduced from the five-year Wilkinson Microwave Anisotropy Probe (WMAP) data1, A(7Li) = 2.72 ± 0.06 (95% confidence level; ref. 3), although this leaves little room for the post-BBN chemical evolution23,24, that is, the contribution of freshly synthesized Li to the ISM by stellar and cosmic ray nucleosynthesis (see representative models23 in Fig. 2). Our estimate of A(7Li)SMC is also consistent with the upper envelope of Li abundances in Milky Way thin-disk stars (Fig. 2)22.

Figure 2: Estimates of the lithium abundance in the SMC interstellar medium and in other environments.
figure 2

Our best estimate for the interstellar (gas+dust phase) abundance of A(7Li) in the SMC (red circle) is derived from the 7Li i/K i ratio. The present day metallicity of the SMC from early-type stars is [Fe/H] = −0.59 ± 0.06. (All uncertainties are 1σ.) The point marked BBN and the dotted horizontal line show the primordial abundance predicted by standard BBN3. The green curves show recent models23 for post-BBN 7Li nucleosynthesis due to cosmic rays and stars. By adjusting the yields from low-mass stars, the models are forced to match the Solar System meteoritic abundance21 (see Supplementary Information). The solid and dashed lines correspond to models A and B23, which include (A) or do not include (B) a presumed contribution to 7Li from core-collapse supernovae. The blue hatched area shows the range of abundances derived for Population II stars in the Galactic halo6, with the ‘Spite plateau’ in this sample at A(7Li)Pop II ≈ 2.10 ± 0.10 (ref. 6). The violet hatched area shows the range of measurements seen in Galactic thin-disk stars, and the thicker violet lines denote the six most Li-rich stars in a series of eight metallicity bins22. The selection of thin-disk stars includes objects over a range of masses and temperatures, including stars that are expected to have destroyed a fair fraction of their Li. Thus, the upper envelope of the distribution represents the best estimate of the intrinsic ISM Li abundance at the epoch of formation for those stars, and the thicker hatched area for the thin-disk sample is most appropriate for comparison with the SMC value. The most Li-rich stars in the Milky Way thin disk22 within 0.1 dex of the SMC metallicity give A(7Li)Milky Way = 2.54 ± 0.05, consistent with our estimate of A(7Li)SMC = 2.68 ± 0.16.

PowerPoint slide

Full size image

However, given the uncertainties in scaling to A(7Li)SMC, the stronger result is our measurement of [7Li/K]SMC = +0.04 ± 0.10. We compare [7Li/K]SMC with measurements6,21,22 of [7Li/Fe] and chemical evolution models23 in Fig. 3. The stars show a rapid decrease in [7Li/Fe] with increasing metallicity until [Fe/H] ≈ −1, at which point the Li abundance increases roughly in lockstep with Fe such that disk stars have a nearly constant [7Li/Fe] ratio, similar to that found in the Solar System. Our measurement of the present-day 7Li-to-metal ratio in the SMC is in agreement with the nearly constant values found in the atmospheres of Milky Way disk stars (≲ ≲0), most of which formed over 4 billion years ago, with the Solar System and the modern-day Milky Way ISM16.

Figure 3: Estimates of Li/Fe in the SMC interstellar medium and in several different environments.
figure 3

The SMC value is derived from the 7Li i/K i ratio. At low metallicities (≲), stellar measurements6 trace the build-up of Fe with a constant Li abundance along the Spite plateau. At higher metallicities, disk star abundances22 show a turnover to roughly constant [7Li/Fe] at values consistent with the Solar System meteoritic value21 (shown as the dash-dotted black line at [7Li/Fe] = 0). Our SMC estimate is consistent with the Solar System and disk star abundances in this region of relatively constant 7Li/Fe abundances, with [7Li/Fe]SMC = +0.04 ± 0.14 for [K/Fe]SMC = 0.0 ± 0.10 (Supplementary Information). The most Li-rich disk stars within 0.1 dex of the SMC metallicity have a mean [7Li/Fe] = −0.13 ± 0.05. (All uncertainties are 1σ.) The green curves show the chemical evolution models23 as in Fig. 2, whereas the dotted line shows the behaviour of [7Li/Fe] for the standard BBN primordial abundance with no subsequent evolution of 7Li. The relative uniformity of the stellar 7Li/Fe abundances at ≳ could be caused by a delicate balance of Li and Fe production and metallicity-dependent depletion of the surface Li abundances (not ruled out given the changes in mean age and mass potentially present in the sample22). However, the agreement of the [7Li/Fe] ratio seen in these old stars (ages exceeding 4 billion years22) and in the present-day interstellar medium of the SMC suggests little change in the stellar abundances for metallicities ≈ up to the solar metallicity. To bring the stellar and SMC interstellar abundances into agreement with standard BBN predictions requires a delayed injection of significant 7Li from stellar production mechanisms as well as vigorous depletion of stellar surface 7Li abundances at metallicities just below that of the SMC.

PowerPoint slide

Full size image

Both the thin-disk stars and our SMC measurements are below standard BBN predictions with reasonable assumptions about post-BBN production, although it is often assumed these stars have had significant depletion of their surface Li abundance23. Taken at face value, the consistency of our SMC measurement with the [7Li/Fe] for those stars calls this assumption into question. Although the models in Figs 2 and 3 are imprecise given the uncertain Li yields from stellar sources, they illustrate the tension between standard BBN predictions and our measurements if there is any post-BBN Li production. This tension can be relieved if a metallicity-dependent depletion of Li in stellar atmospheres is fine-tuned in such a way that it is very strong below [Fe/H] ≈ [Fe/H]SMC = −0.6 (to create the Spite plateau and avoid overproducing Li in the SMC ISM) and negligible at or above the SMC metallicity, thus conspiring to create a constant [7Li/Fe] ratio above [Fe/H] ≈ −1. Alternatively, non-standard BBN scenarios can be invoked to allow for a lower primordial Li abundance4,25.

If non-standard Li production occurs in the BBN epoch, many such models predict excess 6Li compared with the standard BBN. The only known source of post-Big Bang 6Li is production via cosmic ray interactions with ISM particles. Excess 6Li at the metallicity of the SMC would support non-standard production mechanisms, either in the BBN epoch10 or through the interaction of pregalactic cosmic rays with intergalactic helium26. Measurements of 6Li in stellar atmospheres are extremely difficult because the stellar line broadening is well in excess of the isotope shift. However, the 7Li i doublet is well separated in our data owing to the very low broadening in the cool ISM probed by Li i absorption. Our best fit to the SMC Li i absorption gives (6Li/7Li)SMC = 0.13 ± 0.05 (see Supplementary Information and Fig. 1), giving a formal limit to the isotopic ratio in the SMC of (6Li/7Li)SMC < 0.28 (3σ). With higher signal-to-noise ratios and resolution it should be possible to lower the limits for the interstellar isotope ratio in the SMC to provide constraints on non-standard BBN models. This approach has the advantage that ionization and dust-depletion effects are not important for comparing the two isotopes of Li (ref. 27), making 6Li/7Li a powerful diagnostic of nucleosynthesis and non-standard evolution of Li abundances.