Back to the Lithium Plateau with the [Fe/H] < −6 Star J0023+0307^∗

The study of most ancient stars in the Milky Way allows us to infer early properties of the Galaxy, its chemical composition, and assembly history. The number of ultra metal-poor stars (i.e., [Fe/H] < −4.0) has increased strikingly in the past few years. A large observational effort has enlarged the sample of such rare objects, thanks to new search techniques (see, e.g., Christlieb et al. 2001; Frebel et al. 2005; Caffau et al. 2011; Keller et al. 2014; Bonifacio et al. 2015; Hansen et al. 2016; Starkenburg et al. 2018). The main sources of metal-poor candidates are large spectroscopic surveys such as the Hamburg/European Southern Observatory (ESO; Christlieb et al. 2001), Sloan Digital Sky Survey (SDSS; York et al. 2000), or Large Sky Area Multi-Object Fibre Spectroscopic Telescope (LAMOST; Deng et al. 2012), and photometric ones like Skymapper (Keller et al. 2007) or Pristine (Starkenburg et al. 2017).

Metal-poor stars are invaluable messengers that carry information from early epochs and an important key to understand the primordial production of lithium and the processes responsible for the possible meltdown (see, e.g., Aoki et al. 2009; Sbordone et al. 2010) of the lithium plateau (Spite & Spite 1982; Rebolo et al. 1988). All of the stars with metallicities below [Fe/H] < −3.0 and lithium abundances lower than A(Li) = 2.2 ± 0.1 are considered to be likely affected by destruction in the stars. Several explanations have been discussed (see, e.g., Richard et al. 2005; Piau et al. 2006; González Hernández et al. 2009; Meléndez et al. 2010; Sbordone et al. 2010; Molaro et al. 2012; Bonifacio et al. 2015; Matsuno et al. 2017), including lithium depletion in stellar atmospheres due atomic diffusion, high rotational velocities during the earlier star formation, astration by Population III stars, destruction in the pre-main-sequence phase, and the presence of enhanced fragmentation in the formation of ultra metal-poor stars.

New or poorly measured nuclear reaction resonances could affect the lithium production predicted in big bang nucleosynthesis (BBN; e.g., Cyburt & Pospelov 2012), although recent experiments make this nuclear fix to the BBN unlikely (O'Malley et al. 2011). Processes injecting neutrons at the relevant temperatures of the primordial plasma can also alter the primordial ⁷Li abundance, for instance, decaying massive particles (well-motivated candidates such as weakly interacting massive particles are provided by supersymmetry) are a natural source for a neutron excess (Peskin 2008), which injected at the appropriate level may reduce the primordial production of ⁷Li. In addition, time-varying fundamental constants may lead to a significantly lower value for primordial ⁷Li, which is particularly sensitive to changes in the deuteron binding energy (see, e.g., Fields 2011). Lithium observations in stars at the lowest observed metallicities are specially important to bring insight on the processes of potential lithium depletion in stars and, ultimately, to establish if any non-standard physics may have played a role during or after BBN.

J0023+0307 was discovered and first observed by Aguado et al. (2018a). François et al. (2018) observed it with X-shooter on the 8.2 m Very Large Telescope (VLT), providing for first time individual abundances of carbon, magnesium, and silicon. Most recently, using the Magellan Inamori Kyocera Echelle (MIKE) on the Magellan telescope, Frebel et al. (2019) added additional elemental abundances and claimed a lithium detection. Here, we measure with high precision many elemental abundances that confirm most of the results from the literature and provide a reliable lithium abundance determination of J0023+0307. These observations allow us to explore the lower-metallicity end of the A(Li)–[Fe/H] diagram.

J0023+0307 (g = 17.90 ± 0.01, R.A. = 00^h23^m1400, decl. = +03°07'5807 J2000) was observed under the ESO program 0101.D-0149(A) with the UVES spectrograph on the 8.2 m Kueyen VLT at Cerro Paranal Observatory, in the Atacama desert, Chile. The observations were taken in 12 observing blocks (OBs), 1 hr each on 2018 July 22–23, August 8–12, and October 7–8. The total on-target integration time was 10 hr. A 14 slit was used with 2 × 2 binning in dark sky conditions. Our settings used dichroic $\#{Dic}1$ and provided a spectral coverage between 390 and 580 nm. The data were reduced by the ESO staff, and retrieved through the Phase 3 query form. We corrected each spectrum for the barycentric velocity, coadded and combined them. The signal-to-noise per 0.018 Å/pixel in the combined spectrum was ∼30, 55, and 80 at 380, 510, and 670 nm, respectively. The nominal resolving power for this set up is R ∼ 31,220 for the blue part of the spectrum (330–452 nm) and R ∼ 31,950 for the red (480–680 nm). The seeing was 091 during the first OB and between 040 and 065 during the rest of the run, thus the actual R value is somewhat higher. We performed a fit using an automated fitting tool based on the IDL MPFIT routine (with continuum location, global shift, abundance, and global FHWM as free parameters) of all individual, isolated, detected spectral lines with a high-resolution model and obtained an average global (including instrumental and macroturbulent broadening and assuming no rotation) Gaussian broadening of 7.4 ± 0.6 km s⁻¹ (equivalent to R ∼ 40,500). From a cross-correlation of the J0023+0307 spectrum with a extremely metal-poor (EMP) F-type star template we derive a radial velocity of v_rad = −195.5 ± 1.0 km s⁻¹, in perfect agreement with the value reported by Frebel et al. (2019), v_rad = −194.6 ± 1.2 km s⁻¹. No signal of v_rad variation has been observed in the observations spanning four months, suggesting that the star may be single. The Gaia DR2 (Gaia Collaboration et al. 2018) provides a parallax of ϖ = 0.27 ± 0.14 mas. A detailed calculation of the Galactocentric orbit has been recently carried out by Sestito et al. (2019), who find that the orbit of J0023+0307 is confined to the Galactic plane within $| Z| \lt 2.3$ kpc.

Aguado et al. (2018a) reported effective temperatures from the analysis of the Baryon Oscillation Spectroscopic Survey (BOSS; T_eff = 6295 ± 36 K), the Optical System for Imaging and low-Intermediate-Resolution Integrated Spectroscopy (OSIRIS; T_eff = 6140 ± 132 K), and the Intermediate-dispersion Spectrograph and Imaging System (ISIS; T_eff = 6188 ± 84 K) low-/mid-resolution spectra. After that, François et al. (2018) derived a value of T_eff = 6160 ± 100 K from the (g–z) color calibrations. By fitting the wings of the H_α Balmer line we find a higher temperature (T_eff = 6400 ± 150 K; see Figure 1). In addition, we used the color calibrations based on the infrared flux method presented in González Hernández & Bonifacio (2009) with $2{MASS}$ and V Johnson magnitudes transformed from UKIDSS J, H, K, and SDSS g, r filters. This approach also led to a higher temperature (T_eff = 6482 ± 224 K). On the other hand, using color–temperature–metallicity calibrations by Casagrande et al. (2010) we arrive at three very different temperatures from different colors, V–J, V–H, and V–K, giving an average of T_eff = 6441 ± 158 K.

Zoom In Zoom Out Reset image size

Finally, Frebel et al. (2019) use a lower value (T_eff = 5997 ± 130 K) from the V–I color. The dispersion found along the several methods performed by different authors reflects how difficult it is to derive a robust effective temperature in EMP stars (see, e.g., Sbordone et al. 2010). We decided to adopt the T_eff value from François et al. (2018), T_eff = 6160 ± 100 K, which is essentially the same as that derived by Aguado et al. (2018a) from a high signal-to-noise ratio (S/N) ISIS spectrum and also very close to the weighted mean temperature of the values given in Table 1.

The surface gravity derived by Aguado et al. (2018a), $\mathrm{log}g=4.9\pm 0.5$, confirms that J0023+0307 is a dwarf star. François et al. (2018) assumed $\mathrm{log}g=4.5\pm 0.5$ arguing that the star is not bright enough to be considered a subgiant based on the distance derived from the Gaia DR2 parallax. In addition, Sestito et al. (2019) carried out a kinematical study and concluded that the subgiant solution implies an unbound orbit, favoring the dwarf against the subgiant solution at an 88% confidence level. Since there are no detectable iron lines in the UVES spectrum of J0023+0307, we adopt the value from Sestito et al. (2019), $\mathrm{log}g=4.6\pm 0.1$, which is consistent with all of the other estimates. Following the method explained in Allende Prieto et al. (2006), based on the comparison of the main stellar parameters with different sets of isochrones from Bertelli et al. (1994), we find a Galactocentric distance of ∼2.6 kpc. The obtained value is in agreement with those derived by Sestito et al. (2019), d = 2.7 ± 1.4 kpc, using Gaia DR2 data.

Finally, the microturbulence has been fixed at 2 km s⁻¹. The errors introduced using this value instead of a value more appropriate for a dwarf of 1.5 km s⁻¹ are small (<0.1 dex), as discussed in Aguado et al. (2018b) and Starkenburg et al. (2018), and included in our quoted derived abundance uncertainties in this work. In the metallicity regime [Fe/H] < −5.0, the analysis is performed with model atmospheres computed with [Fe/H] = −5.0.

A grid of synthetic spectral models has been computed with the code ASSET (Koesterke et al. 2008) using stellar models from Kurucz ATLAS 9 (Mészáros et al. 2012) and following the recipe explained in Allende Prieto et al. (2014) and Aguado et al. (2017). The grid contains models spanning $-7\,\leqslant [\mathrm{Fe}/{\rm{H}}]\leqslant -4;$ $-1\leqslant [{\rm{C}}/\mathrm{Fe}]\leqslant 7$, $4750\,K\leqslant {T}_{\mathrm{eff}}\leqslant 7000\,{\rm{K}}$, and $1.0\leqslant \mathrm{log}g\leqslant 5.0$. This grid extends the one used by Starkenburg et al. (2018) to reach lower metallicities. Both grids have been smoothed to R ∼ 40,500 and normalized. We have performed an MCMC analysis based in self-adaptive randomized subspace sampling (Vrug et al. 2009) implemented in the FERRE code⁵ (Allende Prieto et al. 2006). The MCMC algorithm samples the probability distribution to estimate central values and uncertainties. Further details about the use of FERRE with high-resolution spectroscopy are presented in Starkenburg et al. (2018).

4.1. Iron and Titanium

4.2. Carbon

4.3. Other Elements

J0023+0307 exhibits relatively high abundances of some elements compared to iron or calcium (see Table 3). The spectrum presents strong Mg i absorptions in the blue (∼383 nm) and also in the magnesium triplet Mg ib at 516–518 nm. Six very clear features (see Table 2) have been identified and fitted giving an abundance of A(Mg) = 4.60 ± 0.11 corresponding to [Mg/H] = −2.93. Si also shows an absorption at (390.5 nm) leading us to derive A(Si) = 4.05 ± 0.05 ([Si/H] = −3.46). The shape of the aluminum features at 394.4 and 396.1 nm is not so well defined, however the derived abundances are the same, A(Al) = 2.35 ± 0.08 ([Al/H] = −4.02). The calcium H&K analysis is significantly improved compared to that presented in Aguado et al. (2018a). Both resonance lines are resolved and a complex interstellar medium (ISM) structure is present (see Figure 1, upper left panel). We find A(Ca) = 0.66 ± 0.09 ([Ca/H] = −5.65), which is ∼0.5 dex higher than Aguado et al. (2018a), from a significantly lower resolution (R ∼ 2500) spectrum, which is fully compatible with François et al. (2018) and Frebel et al. (2019). Assuming a [Ca/Fe] ≥ 0.40, we use the calcium abundance to set a metallicity upper limit of [Fe/H] ≤ −6.1. Finally the two sodium lines around ∼589.2 nm present broader line profiles than expected and moderate quality. The best FERRE fits allow us to derive A(Na) = 2.08 ± 0.21 ([Na/H] = −4.09). Some of the derived upper limits are shown in table.

Table 2.  Absolute Abundances for Individual Lines

Element	λ (Å)	$\mathrm{log}{gf}(\mathrm{dex})$	EW(mÅ)	A(X)
Li i	6707.8	⋯	19.3 ± 1.1	2.02
C (CH)	G-band	⋯	⋯	6.20
Na i	5889.95	0.11	9.2 ± 1.0	1.94
Na i	5895.92	−0.18	12.9 ± 0.9	2.22
Mg i	3829.35	−0.23	79.7 ± 2.0	4.66
Mg i	3832.30	0.12	83.3 ± 2.0	4.50
Mg i	3838.29	0.39	92.2 ± 2.0	4.63
Mg i	5167.32	−1.03	63.7 ± 1.2	4.78
Mg i	5172.68	−0.40	89.2 ± 1.4	4.54
Mg i	5183.60	−0.18	94.7 ± 1.2	4.48
Al i	3944.00	−0.62	24.2 ± 1.4	2.36
Al i	3961.52	−0.32	24.3 ± 1.4	2.34
Si i	3905.52	−1.09	36.7 ± 1.8	4.05
Ca ii	3933.66	0.13	92.82 ± 1.9	0.68
Ca ii	3968.47	−0.16	55.9 ± 2.3	0.65
Fe i	3859.92	−0.68	⋯	<2.02

Download table as:  ASCIITypeset image

Table 3.  Abundances for Individual Species J0023+0307

		This Work				François et al. (2018)				Frebel et al. (2019)
Species	$\mathrm{log}{\epsilon }_{\odot }$ (X)a	$\mathrm{log}\epsilon$ (X)	[X/H]a	σ	N	$\mathrm{log}\epsilon$ (X)	[X/H]b	σ	N	$\mathrm{log}\epsilon$ (X)	[X/H]b	σ	N
Li i	1.05	2.02	⋯	0.05	1	⋯	⋯	⋯	⋯	1.70	⋯	0.20	1
CH	8.39	6.20	−2.19	0.20	⋯	6.40	−1.99	0.30	⋯	5.89	−2.50	0.20
Na i	6.17	2.08	−4.09	0.21	2	⋯	⋯	⋯	⋯	2.24	−3.93	0.30	2
Mg i	7.53	4.60	−2.93	0.11	7	5.10	−2.43	0.23	3	4.63	−2.90	0.10	8
Al i	6.37	2.35	−4.02	0.12	2	⋯	⋯	⋯	⋯	2.30	−4.07	0.20	2
Si i	7.51	4.05	−3.46	0.10	1	4.20	−3.31	0.23	1	3.94	−3.57	0.20	1
Ca ii	6.31	0.66	−5.65	0.09	2	0.60	−5.71	0.20	1	0.57	−5.74	0.20	1
Ti ii	4.95	−0.42	−5.37	0.23	15	⋯	⋯	⋯	⋯	<0.6	<−4.35	⋯	⋯
Cr i	5.64	<1.54	<−4.1	⋯	⋯	⋯	⋯	⋯	⋯	<1.4	<−4.24	⋯	⋯
Fe i	7.45	<1.95	<−5.5	⋯	⋯	<3.5	<−3.95	⋯	⋯	<1.9	<−5.55	⋯	⋯
Ni i	6.23	<2.93	<−3.3	⋯	⋯	⋯	⋯	⋯	⋯	<2.2	<−4.03	⋯	⋯
Sr ii	2.92	<−1.68	<−4.6	⋯	⋯	<0.0	<−2.92	⋯	⋯	<−1.5	<−4.42	⋯	⋯
Ba ii	2.17	<−1.33	<−3.5	⋯	⋯	⋯	⋯	⋯	⋯	<−1.3	<−3.47	⋯	⋯

Notes.

^aSolar abundances from Asplund et al. (2005). ^bSolar abundances converted to Asplund et al. (2005).

Download table as:  ASCIITypeset image

4.4. Lithium

We have been able to clearly detect the Li doublet (see Figure 2) in the rebinned spectrum of the extremely iron-poor star J0023+0307. We measure an S/N of 130 with a bin of 0.035 Å/pixel near the Li doublet. We fit the lithium profile with our automated fitting tool (see Section 2), using a grid of model spectra for three different Li abundances (A(Li) = 2.4, 2.0, 1.6 dex), and with A(Li), continuum location and global shift of the line as free parameters, and we fix the global FWHM to 7.4 km s⁻¹ (see Section 2). We derive an EW(Li) = 19.3 ± 1.1 mÅ (see Table 2). To evaluate the statistical uncertainty, we perform a Monte Carlo simulation with 10,000 realizations by injecting noise, corresponding to the actual S/N near the Li line, in the best-fit synthetic spectrum. We measure a Li abundance in the star J0023+0307 of A(Li) = 2.02 ± 0.08, including the error due to the uncertainty of the adopted T_eff (see, e.g., González Hernández et al. 2008). The difference of ∼0.3 dex in the Li abundance with respect to the value reported in Frebel et al. (2019) cannot be simply atributed to our hotter adopted T_eff. A 160 K lower T_eff translates into about a 0.1 dex lower Li abundance, so a 0.2 dex difference still remains to be explained. We believe that our better quality data allow us to provide a robust Li abundance determination.

Zoom In Zoom Out Reset image size

Our Li abundance measurement is almost at the level of the lithium plateau but at metallicity level of [Fe/H] < −6, which is significantly lower than any previous measurement. In Figure 2 we also display the lithium abundances of extremely metal-poor unevolved stars including main-sequence and turn-off field stars with available Li measurements or upper limits. Matsuno et al. (2017) suggested that no stars with [Fe/H] < −3.5 show Li abundances at the level of the lithium plateau (A(Li) = 2.2 ± 0.1), apart from three stars (González Hernández et al. 2008; Bonifacio et al. 2018). Recently, Bonifacio et al. (2018) have been able to measure the Li abundance of the iron-poor star J1035+0641 (A(Li) = 1.9, [Fe/H] < −5.2). Both stars are suggested to be CEMP-no stars, but J0023+0307 has a lower C abundance of roughly 1 dex. Other unevolved CEMP stars at [Fe/H] < −4.5 have either no Li detected or A(Li)  < 1.8, which may suggest that Li is not connected with the C content (Matsuno et al. 2017).

J0023+0307 is a dwarf star with [Fe/H] < −6.1, A(Li) = 2.02 and very large overabundances of carbon ([C/Fe] > 3.9), magnesium ([Mg/Fe] > 3.2), and silicon ([Si/Fe] > 2.6). These chemical abundance ratios provide a strong argument against ISM pollution by a pair-instability supernova (PISN), since similar shortages of Ca–Fe and Mg–Si should be expected. In particular Takahashi et al. (2018) found the ratios [Na/Mg] ∼ 1.5 and [Ca/Mg] ∼ 0.5–1.3 as the most robust discriminant of PISN. For J0023+0307, both [Na/Mg] = −1.1 and [Ca/Mg] = −3.2 are quite far from those values. The low Fe abundance of J0023+0307 is fully compatible with a Population III supernova scenario. The high [C/Fe] together with Mg and Si derived abundances may point to fallback (Umeda & Nomoto 2003), returning Fe and Ca to the black hole. J0023+0307 is in the low band of the absolute carbon abundance (Bonifacio et al. 2015; Yoon et al. 2016) and does not show radial velocity variations, discarding the scenario of mass transfer from a binary companion. Hartwig et al. (2018) present complementary criteria to disentangle different second generation star formation scenarios, mono-enriched or multi-enriched. The proposed methodology uses two different criteria, a semi-analytical model to determine the formation sites and the other based on the divergence of the chemical displacement (DCD), for further details see Hartwig et al. (2018, 2019). In particular, the chemical signatures of J0023+0307 may fit well in the most likely mono-enriched area of the [Mg/C] versus [Fe/H] diagram presented in Hartwig et al. (2018). SDSS J1035+0641 (with a metallicity of [Fe/H] < −5.2) discovered by Bonifacio et al. (2015) also presents a high probability of being a second generation mono-enriched star (Hartwig et al. 2019).

Bonifacio et al. (2018) has recently detected lithium (A(Li) = 1.9) in J1035+0641 close to the lithium plateau. J0023+0307 with an Li abundance of A(Li) = 2.02 surprisingly nearly recovers the same level of the lithium plateau at about 1 dex less iron content. The presence of lithium in this extremely iron-poor star at [Fe/H] ≲ −6 reinforces the production of lithium at the big bang and places a stringent constraint to any theory aiming at explaining the cosmological Li problem. The fact that no star in this large metallicity regime (−6 < [Fe/H] < −2.5) has been detected to show an Li abundance between that inferred from SBBN and the lithium plateau makes this upper boundary of the Li abundance (or extended Li plateau) at low metallicities difficult to explain by destruction in the stars, and may support a lower primordial Li production, driven by non-standard nucleosynthesis processes.

D.A. thanks the Leverhulme Trust for financial support. D.A. acknowledges the Spanish Ministry of Economy and Competitiveness (MINECO) for the financial support received in the form of a Severo-Ochoa PhD fellowship, within the Severo-Ochoa International PhD Program. D.A., J.I.G.H., C.A.P., and R.R. also acknowledge the Spanish ministry project MINECO AYA2017-86389-P. J.I.G.H. acknowledges financial support from the Spanish Ministry of Economy and Competitiveness (MINECO) under the 2013 Ramón y Cajal program MINECO RYC-2013-14875.