betaFeedback

WorksPhysics and Astronomy

Unexpected Plateau: The Constant log⁡ε(Th/Eu) \rm \log ε(Th/Eu) Ratio of Actinide-Boost Stars

Abstract

The astrophysical site of the rapid neutron-capture process (rr-process) is still one of the major open questions in astronomy. A key mystery is the origin of the peculiar actinide-to-lanthanide ratio found in the so called actinide-boost stars, with log⁡ε(Th/Eu)>−0.35 \rm \log ε(Th/Eu)>-0.35. Here we present the largest sample of Th measurements in stars formed in an external galaxy, Gaia Sausage Enceladus. From this sample, we present the three most metal-rich actinide-boost stars discovered to date, −1.6<[Fe/H]<−1.1\rm -1.6<[Fe/H]<-1.1. In addition, we measured Th in a comparison sample of Milky Way stars formed in situ, and compiled Th measurements from the literature which do not account for non-local thermodynamic equilibrium (NLTE) effects. After applying NLTE corrections to the combined sample, a bimodality in the log⁡ε(Th/Eu)\rm\logε(Th/Eu) ratio is revealed. Unexpectedly, actinide-boost stars lie on a very narrow (σ=0.02)(σ=0.02) plateau of log⁡ε(Th/Eu)=−0.2\rm \logε(Th/Eu)=-0.2, clearly separated from actinide-normal stars. Most (if not all) of the actinide-boost stars in our sample originate from outside the Milky Way. Chemical evolution models show that such a plateau can arise, even at [Fe/H] = −-1, when a single actinide-rich event dominates the r-process content of a small or gas-depleted system. Combined with other evidence, our results suggest that mergers with neutron stars are the most promising astrophysical site for actinide-boost stars.

Introduction

The astrophysical site of the rapid neutron-capture process (rr-process) is one of the most persistent open questions in astronomy (e.g. [108]). It has long been established that elements heavier than Zn (Z>30Z > 30) are primarily created through a series of neutron-captures, and consequent β\beta-decays (Burbidge et al. [10]). Around half of the abundances of these heavy elements come from the rr-process, yet its origin remains uncertain.

Several sites have been proposed to have the physical conditions for sufficiently high neutron-flux. The favoured sites for the rr-process have long been mergers including neutron stars (e.g. [70]), and the detailed observations of GW170817 showed that a merger of two neutron stars (NSM) can indeed host the rr-process (e.g. [90, 127]). In addition, sites associated with massive stars have also been proposed. These include rare classes of core-collapse supernovae (ccSN), such as magneto-rotational supernovae (mrSN) or collapsars ([84, 83, 111, 96, 95]); and magnetar giant flares ([15, 88 b]). Currently, none of these prompt rr-process sites are confirmed observationally, making their role in the creation of rr-process elements very uncertain.

A remarkable property of the rr-process is that relative abundances from the second peak through the lanthanides (56≤Z≤7156 \leq Z \leq71) appear to be universal, from the most metal-poor stars (e.g. [119]) to the Sun (e.g. [11]). However, this robustness breaks down at the low-mass (Z<56Z < 56) and high-mass (76≤Z≤9276 \leq Z \leq92) ends (e.g. Hill et al. 2022). In particular, the so-called actinide-boost stars show an enhancement of log⁡ϵ(Th/Eu)>−0.35\log\epsilon(\mathrm{Th}/\mathrm{Eu}) > -0.35 (e.g. [54, 71]). The underlying cause of these high actinide abundances is still unknown, but might suggest a different nucleosynthetic site (e.g. NSM vs. mrSN) and/or different physical conditions (e.g. [55, 125] (Eichler et al. 2019)).

One of the most promising ways of gaining insights into the rr-process is through Galactic Archaeology. The chemical evolution of the rr-process elements can be followed by observations of individual stars in the Milky Way and its satellite systems. These studies suggest that at least two rr-process sites are needed to explain the data: a delayed site (typically associated with NSM) and a prompt source, whose origin is still debated [117, 82, 66] (Côté et al. 2019). Dynamical studies have associated rr-rich stars with groups that are remnants of accreted dwarf galaxies [99, 131, 40, 48]. These stars have also been found in several dwarf galaxies, ranging from ultra-faint dwarfs (e.g. [47] (Ji et al. 2016)) to large dwarf spheroidals such as Fornax [94, 74]. Going towards even more massive galaxies, Ernandes et al. [25] found 10 rr-II stars ([Eu/Fe] >> 1.0 and [Ba/Eu] << 0; [7]) that are kinematically part of the Gaia Sausage Enceladus (GSE), which has been dubbed the last major merger of the Milky Way, occurring ∼10\sim10 Gyr ago [86, 50]. Consistent with this picture, a cosmological simulation of a Milky Way analogue finds that most r-II stars originate from dwarf galaxies that are later accreted. The predicted accreted fraction is 100% at [Fe/H] =−3= -3 and 97% at [Fe/H] =−2= -2 and it decreases with metallicity [52, 53].

Among the rr-process enhanced stars, about 30% present actinide-boost, with an over-abundance of [Th/Eu] [57]. Lin et al. [71] reported the first confirmed actinide-boost star in GSE, with log⁡ϵ(Th/Eu)=−0.22\log\epsilon(\mathrm{Th}/\mathrm{Eu}) = -0.22, and found that a magnetorotationally driven jet-supernova model best reproduces its abundance pattern. Their kinematic analysis also indicates that approximately two-thirds of the actinide-boost stars they consider are ex situ. Together, these results suggest that low-mass accreted systems may provide conditions in which a rare actinide-boost rr-process event can strongly enrich a poorly mixed gas reservoir.

In this paper, we measure the actinide-to-lanthanide ratio in 23 stars formed in GSE, and in 27 Milky Way stars formed in situ. Thus, we present the first systematic study of the actinide-to-lanthanide ratio in stars from an external galaxy.

Stellar sample

Primary sample

The primary sample used in this study is a sub-sample of 50 stars selected from Nissen & Schuster [86, 87] and Nissen et al. [85], NS sample hereafter. The stellar sample consists of nine thick disk stars and 41 halo stars (Vtot>180V_{\mathrm{tot}} > 180 km s−1^{-1}) encompassing both high- and low-α\alpha populations. The high-resolution (R≥40,000R \ge40{,}000) and high S/NS/N (≥80\ge80 pix−1^{-1} at 4000 Å) optical spectroscopic data for 44 stars in the sample come from new observations via two ESO VLT/UVES programs (PI: Skúladóttir, ESO IDs 109.22 and 110.240). For the remaining six stars, raw echelle spectra of comparable quality were adopted from the UVES archive2. These spectra were acquired with the same UVES instrumental configurations as our observations, ensuring structural homogeneity across this dataset.

All spectra were reduced using the standard UVES pipeline [33]. Individual exposure frames were Doppler-corrected, co-added to maximize S/NS/N, and carefully continuum-normalized. The detailed procedure of data processing is discussed in our previous paper [26]. We have directly adopted the high-precision stellar atmospheric parameters and Mg abundances of the program stars from Ernandes et al. [26] that originally come from Nissen & Schuster [87] and Nissen et al. [85]. This sample consists of unevolved dwarf stars (log⁡g>3.8\log g > 3.8) with metallicities in the range −1.62<[Fe/H]<−0.44-1.62 < [\mathrm{Fe}/\mathrm{H}] < -0.44.

Literature sample

To expand our statistical leverage across a wider metallicity range and test the broader universality of the thorium-to-europium ratio, we have compiled from the literature an additional sample of 60 stars (primarily giants with log⁡g≤2.9\log g \le2.9 and metallicity down to [Fe/H] =−3.5= -3.5) in the Milky Way with reported high-resolution spectroscopic abundances for Eu and Th. Only stars with detections are selected and the ones with upper limits are excluded. We have also adopted the Mg abundances as representative of the α\alpha-elements for this sample. To ensure consistency, we made sure that these stars were selected exclusively from studies employing line parameters and methodologies similar to our program star analysis, establishing a homogeneous 110-star baseline with a metallicity range −3.5≤[Fe/H]≤−0.44-3.5 \le[\mathrm{Fe}/\mathrm{H}] \le-0.44.

Analysis

Abundance determination

In this work, we present new measurements of Eu and Th for the NS sample. The elemental abundances are computed with the MOOG code (Sneden 1973; Sneden et al. 2012; version 2019), using ATLAS9 Kurucz grid of model atmospheres [14], assuming Local Thermodynamic Equilibrium (LTE) and 1D. The line parameters are taken from the linemake3 atomic and molecular line database [93].

The europium abundances are derived from the spectral synthesis of the Eu II 4129.725 and 4205.042 Å lines. The hyperfine structures (hfs) and isotopic components are included in the analysis, assuming the solar isotopic ratio 151Eu:153Eu=47.8:52.2{}^{151}\mathrm{Eu} : {}^{153}\mathrm{Eu} = 47.8 : 52.2 [1].

The abundances of Th are derived from the spectral synthesis of the dominant Th II 4019.129 Å feature. This is the only Th line that is strong enough to be detected in our spectra. This spectral feature belongs to a crowded region and is blended with overlapping features of other elements. To isolate the pure Th II contribution, these blends were carefully identified and calibrated using the solar atlas and synthetic spectrum comparisons.

Abundance ratios are calculated for both program sample and the literature sample by adopting the solar abundance values from Asplund et al. [3]. The spectral synthesis fits for the Eu and Th lines of interest for a metal-poor and a metal-rich star in our primary sample are shown in Figure 1. The atmospheric parameters and abundances of the primary and literature sample are given in Tables A.1 and A.2, respectively.

Figure 1. Spectral synthesis fits and the derived abundances for Th and Eu for a metal-poor (HD 132475, [Fe/H] = −1.5; upper panel) and a metal-rich (BD+11 2369, [Fe/H] = −0.5; lower panel) star in our sample. Observed spectra are shown as black dots, along with the best fits (blue lines) and ±0.2 dex deviations (shaded region). Dashed lines show the absence of the element of interest.

Uncertainty in the abundances

The typical uncertainties in the adopted atmospheric parameters are Δ(Teff,log⁡g,vmic,[Fe/H])≃±(50 K,0.10 dex,0.10 km s−1,0.05 dex)\Delta(T_{\mathrm{eff}}, \log g, v_{\mathrm{mic}}, [\mathrm{Fe}/\mathrm{H}]) \simeq\pm(50\ \mathrm{K}, 0.10\ \mathrm{dex}, 0.10\ \mathrm{km\ s}^{-1}, 0.05\ \mathrm{dex}) [86]. Given the high quality of our spectra, these errors are the dominant source of the abundance uncertainty. The deviations in the elemental abundances from the best-fit values are estimated by changing the atmospheric parameters individually by their respective uncertainties. The total uncertainty in the derived abundances is then estimated by adding these contributions in quadrature as explained in Shejeelammal et al. [110], resulting in a typical uncertainty of ∼0.05 dex in log⁡ϵ(X)\log\epsilon(\mathrm{X}) for both Eu and Th. An uncertainty of ∼0.04 dex in [Mg/Fe] is adopted from Nissen & Schuster [86].

Non-LTE corrections

Our combined data set spans two distinct evolutionary stages - dwarfs in our program sample and giants in the literature sample - and most of them have metallicities [Fe/H] < −0.8 (≈ 90%). Applying NLTE correction (ΔNLTE\Delta_{\mathrm{NLTE}}) across our combined sample of 110 metal-poor stars is essential to eliminate evolutionary offsets and to ensure abundance trends reflect true physical yield variations. Even though several 1D/3D NLTE studies are available for various Eu II lines [e.g. 79, 121, 43], Mashonkina et al. [80] is the only NLTE study (1D) available for Th II. Hence, for consistency, 1D NLTE corrections are employed in this study, as described below.

Eu II: The NLTE corrections for the Eu II 4129.725 and 4205.042 Å lines in our primary NS sample are adopted from Storm et al. [122]. Both lines have identical corrections. Since the literature sample comes from multiple high-resolution spectroscopic studies, the reported mean LTE abundances were derived from various Eu II lines (predominantly, Eu II 3907.107, 4129.725, 4205.042, 6437.640, and 6645.064 Å). To calculate the Eu II NLTE corrections of individual stars in this literature sample, we have used the online tool NLiTE Koutsouridou et al. [68]4, optimized for metal-poor stars using state-of-the-art publicly available NLTE grids, in this case from Mashonkina & Gehren [79]. The ΔNLTE(Eu)\Delta_{\mathrm{NLTE}}(\mathrm{Eu}) of the combined sample ranges from +0.06 to +0.27 dex (Tables A.1 - A.2).

Th II: The NLTE corrections for all 110 stars are adopted from Mashonkina et al. [80]. All literature studies are based on LTE abundances of Th. Even though a ΔNLTE\Delta_{\mathrm{NLTE}} is negligible (∼0.01 dex) for dwarfs, it becomes significant at lower surface gravities for a given temperature, reaching up to +0.21 dex (Th II 4019.129 Å) for the TeffT_{\mathrm{eff}} / log⁡g\log g / [Fe/H] = 4000 K/0.0/−2 model [80] (Tables A.1 - A.2). Depending on the stellar parameters, especially log⁡g\log g and [Fe/H], the ΔNLTE\Delta_{\mathrm{NLTE}} has varying impact on the thorium-to-europium ratio. For instance, extremely metal-poor (EMP) evolved giants show similar corrections for both Eu and Th, and hence the ratio remains almost unaffected (e.g. HD 108577, 2MASS J20050670−3057445). Meanwhile the Eu correction is higher compared to Th for dwarfs (Table A.1) and less-evolved EMP giants (e.g. HD 10837), which leads to non-negligible corrections of the ratio. In the combined sample, ΔNLTE(Th/Eu)\Delta_{\mathrm{NLTE}}(\mathrm{Th/Eu}) ranges from −0.19 to +0.07 dex.

Astrometric and Kinematic data

In order to better trace the dynamical origin of our sample, we adopted Gaia astrometric parameters for the whole sample. For that we query the Gaia DR3 IDs using the ASTROQUERY.SIMBAD (Ginsburg et al. [38]) Python package where we got 108 stars that have parallaxes and radial velocity from Gaia Radial Velocity Spectrometer (RVS). Based on the Gaia astrometric data from Gaia DR3 (Gaia Collaboration et al. [35, 36]) and the zero-point corrected parallaxes, using the prescription by Lindegren et al. [73], we calculated the kinematics and orbital parameters (Table A.3) using the Python packages ASTROPY (Astropy Collaboration et al. [6, 5, 4]) and GALPY5 (Bovy [8]).

The kinematic of both the primary and literature sample are shown in Figure 2. We followed the classification of Ernandes et al. [26] to separate the halo stars in the NS sample into in-situ (blue) and GSE (red) populations. For the homogeneity of the literature sample, here the GSE stars are selected based on the high-purity kinematic selection criteria given by Feuillet et al. [28]. GSE stars are shown in red throughout this paper, unless stated otherwise.

Figure 2. Orbital properties of the primary sample and literature sample in different dynamical spaces. Top left panel: Toomre diagram showing the Galactic velocity components, where the dashed curve separates the disk and the halo regions. Middle left panel: Angular momentum LzL_z as a function of orbital eccentricity. Bottom left panel: Angular momentum as a function of the maximum height above the Galactic plane Zmax⁡Z_{\max}. Right panel: Orbital energy as a function of angular momentum. Red circles show the GSE stars, blue open circles in-situ halo stars, black triangles thick disk stars, teal open squares denote the literature sample, and star symbols are the actinide-boost stars.

As shown in Figure 2, all actinide-boost stars in the GSE sample belong to the high-energy class defined by Skúladóttir et al. [114], with E>−4.5×104 km2s−2E > -4.5 \times10^{4}\ \mathrm{km}^{2}\mathrm{s}^{-2}. According to their interpretation, these stars are therefore more likely to have formed in the outer regions of GSE, rather than the more efficiently star-forming inner regions.

Most of the actinide-boost stars in the literature have been associated to an accreted system (e.g. Gudin et al. [40]). More generally, at the lowest metallicities ([Fe/H]<−2.5[\mathrm{Fe}/\mathrm{H}] < -2.5) the large majority of stars in the Milky Way halo is predicted to be of accreted origin (e.g. [67, 53]). We note that three actinide-boost (r-II) stars, 2MASS J14244532-2542471 ([Fe/H]=−3.39[\mathrm{Fe}/\mathrm{H}] = -3.39), BPS CS 31078-018 ([Fe/H]=−2.85[\mathrm{Fe}/\mathrm{H}] = -2.85), and BPS CS 30306-132 ([Fe/H]=−2.42[\mathrm{Fe}/\mathrm{H}] = -2.42), belong to the disk region in the Toomre diagram, and are in general consistent with disk-like kinematics (see Figure 2). However, two of these stars, BPS CS 31078-018 and BPS CS 30306-132, have been identified as belonging to the dynamically tagged groups of Hattori et al. [48], that are ideal candidates for remnants of disrupted r-process-enhanced dwarf galaxies that merged with the ancient Milky Way. The third star with disk-like kinematics, 2MASS J14244532-2542471, is an EMP star and therefore likely to have an ex-situ origin; however, the original paper was not able to associate it to any known structure, so it remains possible that this is an in-situ-formed star (Placco et al. [93]). The other actinide-boost stars in the literature sample exhibit halo kinematics and several previous studies found them to be members of dynamically tagged groups that have likely accreted origin (e.g. [131, 48]).

Abundances of Eu and Th

We identified three actinide-boost (log⁡ϵ(Th/Eu)>−0.35\log\epsilon(\mathrm{Th}/\mathrm{Eu}) > -0.35) stars in our NS sample: G 112−43, HD 284248, and G 87−13 with log⁡(Th/Eu)NLTE\log(\mathrm{Th}/\mathrm{Eu})_{\mathrm{NLTE}} values −0.17, −0.19, and −0.21 dex, respectively. These are the most metal-rich ([Fe/H]>−1.6[\mathrm{Fe}/\mathrm{H}] > -1.6) actinide-boost stars discovered to date, and the only main-sequence stars identified to belong to this class. One of these actinide-boost stars, HD 284248, has available spectra from the Hubble Space Telescope in the ultra-violet. The analysis of this spectra shows that this actinide-boost star also has unusual abundances in other heavy elements (Z>70Z > 70), but this will be discussed further in Singh et al. in prep.

Abundance ratios of [Eu/Fe] and [Th/Fe]

Figure 3 shows the distribution of [Eu/Fe]NLTE_{\mathrm{NLTE}} and [Th/Fe]NLTE_{\mathrm{NLTE}} for our sample. At low metallicities, the scatter is very large with a spread of over 2 dex, while at higher metallicities the stars converge. This behavior of the n-capture elements is well known in the literature (e.g. [e.g. 30, 16]).

Figure 3. Observed [Eu/Fe] and [Th/Fe] ratios of the combined sample with respect to [Fe/H]. Symbols as in Figure 2.

However, the GSE stars in our primary sample ([Fe/H] ≳\gtrsim −1.5, red circles) are on average higher both in [Eu/Fe]NLTE_{\mathrm{NLTE}} (see also e.g. [e.g. 29]) and [Th/Fe]NLTE_{\mathrm{NLTE}} compared to the Milky Way in-situ halo and thick disk stars. In particular, we note that all three actinide-boost stars in our primary sample belong to GSE (3/23: 13−7+1313^{+13}_{-7}%), while none are found among the Milky Way stars selected in the same way (0/27: 0+70^{+7}%). Based on the literature sample, the fraction at low-metallicities, −3 < [Fe/H] < −2, is 7/43: 16−6+916^{+9}_{-6}% (uncertainties estimated using Gehrels [37]).

Finally we note that actinide-boost stars (star symbols in Figure 3), are not necessarily r-process-rich, that is high in [Eu/Fe], but instead follow a distribution very similar to actinide-normal stars.

Bimodality in the (Th/Eu) ratio

The log ϵ\epsilon(Th/Eu) ratios of the sample in LTE and NLTE are shown in Figure 4. When evaluated under LTE assumption, the ratios of the sample exhibit a broad, continuous distribution with a mean value at log ϵ\epsilon(Th/Eu) = −0.43 (Figure 4, upper panel).

Figure 4. Distribution of log⁡ϵ(Th/Eu)\log\epsilon(\mathrm{Th}/\mathrm{Eu}) with respect to [Fe/H], assuming LTE (top) and NLTE (bottom). Dotted lines at log⁡ϵ(Th/Eu)=−0.35\log\epsilon(\mathrm{Th}/\mathrm{Eu}) = -0.35 represent the lower boundary of actinide-boost region. Probability distribution functions are shown on the right side panel, both the observed (solid) and intrinsic (dashed lines). The mean values are shown as horizontal lines (see labels), along with the 1σ1\sigma as shaded regions in the main panels.

However, when the NLTE corrections are applied, the picture changes substantially. Many of the stars previously categorized as actinide-boost stars are now revealed to be actinide-normal. This is especially true for the sample of metal-rich main-sequence stars, where 17 out of the 20 actinide boost-stars in LTE are actually actinide-normal in NLTE. On the other hand, only 1 star out of the metal-poor giants ([Fe/H] < −2) changes from actinide-boost to actinide-normal. Furthermore, the difference seen in LTE between main sequence stars, ⟨log⁡ϵ(Th/Eu)⟩MS\langle\log\epsilon(\mathrm{Th/Eu})\rangle_{\mathrm{MS}} = −0.36, and giant stars, ⟨log⁡ϵ(Th/Eu)⟩giants\langle\log\epsilon(\mathrm{Th/Eu})\rangle_{\mathrm{giants}} = −0.48, disappears when the NLTE corrections are applied.

Thanks to the more accurate NLTE abundances and our three newly discovered actinide-boost stars at high metallicity ([Fe/H] > −1.6), the combined sample is split into two distinct populations, separated by a clean observational gap (Figure 4, lower panel):

  1. Actinide-boost stars: Of the thirteen stars in our sample, twelve belong to a very narrow plateau with a constant value of log ϵ\epsilon(Th/Eu) ≈\approx −0.20, and very low σ∼\sigma\sim 0.02. The only star that deviates from this plateau is 2MASS J14244532-2542471 (J1424 hereafter; [91]).

  1. Actinide-normal stars: Most of the stars (N = 97) in the sample belong to this broader population centered at log ϵ\epsilon(Th/Eu) = −0.54 and σ\sigma = 0.11.

  1. Observational gap: The NLTE corrections lead to a clean observational gap between −0.35 < log ϵ\epsilon(Th/Eu) < −0.25, where no stars in our sample lie.

The plateau of actinide-boost stars and the observational gap persist throughout the wide metallicity range of the sample, −3.5 < [Fe/H] < −1. This indicates that the bimodality of the log ϵ\epsilon(Th/Eu) ratio is a genuine and robust physical feature.

Statistical confirmation

To verify this observed bimodality, we performed a two-stage statistical analysis, as described in detail in Appendix B6. As a first step, a Kolmogorov–Smirnov (KS) test was conducted, which rejected the null hypothesis of unimodality (p<0.05p < 0.05). Gaussian Mixture Models (GMM) were used to further analyze the distributions, indicating that the two-component model for log ϵ\epsilon(Th/Eu)NLTE_{\mathrm{NLTE}} is optimal, while a unimodal distribution is favored for LTE. The NLTE analysis revealed two Gaussian distributions: a standard peak at μ1\mu_1 = −0.54 (σ1\sigma_1 = 0.11) comprising 89.1% of the sample and an actinide-boost peak at μ2\mu_2 = −0.20 (σ2\sigma_2 = 0.02) comprising 10.9%. The LTE distribution is a single Gaussian centered at μ\mu = −0.43 with σ\sigma = 0.18 (see solid lines in the side panel of Figure 4).

A Bayesian Markov Chain Monte Carlo (MCMC) analysis provided intrinsic parameters for these peaks and confirmed a distinct observational gap, indicating a bimodal distribution. The intrinsic parameters for NLTE peaks based on the MCMC analysis are μ1\mu_1 = −0.54±0.01, σint,1\sigma_{\mathrm{int},1} = 0.08±0.01, and μ2\mu_2 = −0.20±0.02, σint,2\sigma_{\mathrm{int},2} = 0.00±0.02. The analysis verified a significant observational gap (−0.35 < log ϵ\epsilon(Th/Eu)NLTE_{\mathrm{NLTE}} < −0.25) between the peaks, with a confidence level of 94.6% (1.9σ\sigma).

The actinide-boost peak is therefore tightly constrained with a narrow observational width (∼\sim0.02 dex) and consistent with zero intrinsic width. The intrinsic mean and width of the LTE distribution are μ\mu = −0.43±0.02 and σint\sigma_{\mathrm{int}} = 0.16±0.01 (Figure 4, side panel dashed lines).

This statistical analysis confirms that the log ϵ\epsilon(Th/Eu) distribution in metal-poor stars is intrinsically bimodal, pointing to distinct astrophysical sites for the r-process in the early Galactic chemical evolution.

Origin of actinide-boost stars

The clear separation in log ϵ\epsilon(Th/Eu)NLTE_{\mathrm{NLTE}} (Figure 4) favors the interpretation that actinide-boost and actinide-normal stars are enriched by two different rr-process sites (see also § 7), rather than by a single site with continuously varying physical conditions. Furthermore, the very narrow range of the plateau at log⁡ϵ(Th/Eu)≈−0.20\log\epsilon(\mathrm{Th}/\mathrm{Eu}) \approx-0.20 indicates that these stars are tracing the intrinsic yields of this site, rather than a mixture of several sites. Two key questions therefore arise: 1) What is the site that produces the actinide-boost stars? 2) What kind of chemical enrichment scenario can explain their formation over such a large metallicity range?

Regarding the first question, theoretical yields predict that NSMs produce higher actinide-to-lanthanide ratios (e.g. Wanajo et al. [126]; Holmbeck et al. [56]; Ricigliano et al. [97]), a result that also holds for BH–NS mergers (Wanajo et al. [125]), than other proposed rr-process sites. These include mrSNe (e.g. Nishimura et al. [84]; Reichert et al. [96, 95]), collapsars (e.g. Miller et al. [81]; Just et al. [63]; Fujibayashi et al. [34], but see Siegel et al. [111]), and magnetar giant flares (Patel et al. [89]).

Regarding the second question, the large scatter of [Eu/Fe] at [Fe/H] << −2-2 (e.g. Figure 3) points to the rr-process being a rare event. At very low metallicity, it is therefore plausible that actinide-boost stars formed in an environment that had experienced only one rr-process event, with boosted actinide production. At higher metallicity, these stars are more challenging to explain. It is hard to imagine an environment being enriched to [Fe/H] ≳−1.5\gtrsim-1.5 by only one rr-process event. Indeed, no stars at this metallicity have been identified as being very rr-process poor. The most likely scenario is thus that, although the birth environment of these stars experienced several rr-process events, their

rr-process abundances are dominated by a single event with high actinide production. In the following, we explore this scenario, assuming that the dominant event is a NSM.

Chemical evolution model

Figure 5 shows the results of our single-zone chemical evolution models in the [Th/Eu], [Eu/Mg], and [Th/Mg] vs [Mg/H] planes, and in the [Th/H] vs [Eu/H] plane. The models represent average trends and do not include any intrinsic scatter. In all models, the radioactive decay of 232Th{}^{232}\mathrm{Th} is included in the evolution of the gas. The decay that takes place in the stars after their formation is shown separately as shaded areas in Figure 5. Following the theoretical predictions mentioned above, we assume that NSM produce the high actinide-to-lanthanide ratio of the plateau. We note, however, that any nucleosynthetic site with the same yields would produce similar results. The [Th/Eu] ratio depends only on the yields of the source that dominates the rr-process budget. This degeneracy makes the plateau relatively easy to fit, and motivates some caution when interpreting the nature of the rr-process source.

Chemical evolution model of the formation of actinide-boost stars

Figure 5. Chemical evolution model of the formation of actinide-boost stars (star symbols). Actinide-normal stars are shown as gray dots. At low metallicity ([Fe/H]<−1.5[\mathrm{Fe}/\mathrm{H}] < -1.5, purple), such stars are formed in small gas clouds enriched by only one rr-process event - a NSM boosted in actinides. Actinide-boost stars at [Fe/H]>−1.5[\mathrm{Fe}/\mathrm{H}] > -1.5 (yellow, this work) are formed when a NSM occurs in regions that have experienced gas depletion. Lines show the model predictions at the time of star formation; shaded areas extend these down to the present-day values, accounting for the radioactive decay of Th in the stars.

The [Eu/Mg] and [Th/Mg] ratios, by contrast, compare an rr-process element to the α\alpha-element budget produced by ccSNe, and therefore also depend on the absolute rr-process mass ejected relative to the gas it enriches. A single rr-process event diluted in a large gas reservoir leaves a too weak imprint and the model will then underproduce the [Eu/Mg] and [Th/Mg] ratios of the actinide-boost stars, even while matching their [Th/Eu]. Reproducing all three ratios simultaneously thus requires a sufficiently high rr-process mass per unit gas mass, that is, either a large number of events or a small gas reservoir.

General chemical enrichment

The actinide-normal stars are reproduced here by adopting the chemical evolution model of GSE from Vincenzo et al. [124] as a reference (grey dashed lines in Figure 5). In this model, the galaxy forms at high redshift by infall of gas from the intergalactic medium into its potential well, at a rate I(t)∝exp⁡(−t/τinf)I(t) \propto\exp(-t/\tau_{\mathrm{inf}}), with an infall timescale τinf=0.24\tau_{\mathrm{inf}} = 0.24 Gyr, normalized so that the total accreted gas mass is 1010 M⊙10^{10}\ M_{\odot}. Stars form following a Schmidt–Kennicutt law, Ψ(t)∝νMgas1.5\Psi(t) \propto\nu M_{\mathrm{gas}}^{1.5}, with a star-formation efficiency ν=0.42 Gyr−1\nu= 0.42\ \mathrm{Gyr}^{-1}. The galaxy suffers galactic winds, which develop when the thermal energy of the gas, heated by stellar winds and SNe, exceeds its binding energy (see Bradamante et al. [9]); Vincenzo et al. [124] found the efficiency of these outflows to be low.

For the rr-process, the model includes two sources with different characteristic timescales: a delayed source, NSMs, described by a delay-time distribution (adopted from Simonetti et al. [112]) with yields from Ricigliano et al. [97] (their model mS∗m_{\mathrm{S}}^{*}), and a prompt source, here represented by collapsars with nucleosynthesis prescriptions from Siegel et al. [111]. Mg is produced by ccSNe, following François et al. [31]. This model reproduces the average trends of actinide-normal stars in all panels of Figure 5 but, as expected, underestimates the Th abundances of the actinide-boost stars.

Actinide-boost enrichment

Although NSMs are predicted to produce a higher [Th/Eu][\mathrm{Th}/\mathrm{Eu}] ratio than prompt sources such as collapsars or mrSNe (see § 6), they do not dominate the rr-process budget in the general model: when NSMs are parametrised as a delayed source with a relatively low rate, the budget is instead governed by the prompt contribution from rare massive stars (see e.g. Molero et al. [82] for details). For this reason, we model the plateau enrichment as a single NSM event, ejecting 0.1 M⊙0.1\ M_{\odot} of rr-process material. This is at the upper end of current predictions, which typically range from ∼10−3\sim10^{-3} to a few ×10−2 M⊙\times10^{-2}\ M_{\odot} (e.g., Goriely et al. [39]; Rosswog et al. [105]; Ricigliano et al. [97]), with the more massive configuration approaching values comparable to the one adopted here. We model this condition in two complementary ways, corresponding to the two plateau sub-populations.

The low-metallicity plateau stars (Figure 5, purple) are modeled as forming in a small environment (with a total accreted gas mass ∝105 M⊙\propto10^{5}\ M_{\odot}), whose rr-process enrichment comes from a single, early NSM that is the only rr-process source. Because the gas reservoir is small, the NSM leaves a strong imprint without the gas having to be depleted first (see below). The resulting track lies on the plateau, matching the most metal-poor plateau stars in [Th/Eu][\mathrm{Th}/\mathrm{Eu}], and is in reasonable agreement with the observations in the other abundance planes, especially considering that low Eu and Th abundances are not detectable at this metallicity.

The model for the more metal-rich GSE plateau stars (Figure 5, gold) instead includes two rr-process sources. A prompt source (e.g., collapsars), active throughout the evolution, enriches the gas as in the general model (§ 6.1.1), and a single, delayed NSM is superimposed on this evolution. The host environment, a region within GSE, has a total accreted gas mass of 107 M⊙10^{7}\ M_{\odot}, assembled on a short timescale (τinf=0.5\tau_{\mathrm{inf}} = 0.5 Gyr), forming stars with a low efficiency (ν=0.04 Gyr−1\nu= 0.04\ \mathrm{Gyr}^{-1}) and subject to a strong galactic wind with a mass-loading factor of ω≃10\omega\simeq10.

With this gas mass, a single NSM would on its own be too diluted to leave a clear imprint. However, the strong wind progressively removes the gas, so that by the time the merger occurs the reservoir is sufficiently depleted for the actinide-rich imprint to become visible. This timing results from the combination of a short infall timescale and the delay of the NSM: the star formation rate peaks early, and the single NSM occurs about 1 Gyr later, by the time the galactic wind has already reduced the gas reservoir by more than half. The NSM then enriches the residual gas at [Mg/H] ≈−1.5\approx-1.5, lifting [Th/Eu] to the plateau value and reproducing the actinide-boost stars, identified by this work. At the same time, the star formation is still high enough to form a new stellar generation. The NSM therefore occurs in a time window in which the gas reservoir is sufficiently depleted for its imprint to be visible, but still gas-rich enough to form the observed population of metal-rich actinide-boost stars.

Finally, we note that although our model reproduces the observed trends, our adopted production value of log⁡ϵ(Th/Eu)\log\epsilon(\mathrm{Th/Eu}) for the actinide-boost rr-process seems to be too low compared to the observations. However, the uncertainty of the theoretical value is very large, and the results presented here can serve as an important observational constraint, see § 7.1.

Discussion

The actinide-boost plateau

As shown in Figure 4, actinide-boost stars lie on a remarkably narrow plateau of log⁡ϵ(Th/Eu)=−0.2\log\epsilon(\mathrm{Th/Eu}) = -0.2 over a large range in metallicity. This discovery was enabled by our new detection of the three most metal-rich actinide-boost stars known to date ([Fe/H] >−1.6> -1.6), and the application of NLTE corrections to our primary and literature samples. This is an indication that the log⁡ϵ(Th/Eu)\log\epsilon(\mathrm{Th/Eu}) ratio of events with boosted actinide production must be extremely robust, since only 1/13 actinide-boost stars falls out of this plateau.

The narrowness of the plateau is very puzzling, as the scatter around the mean value, σ=0.02\sigma= 0.02, is similar to, or even smaller than what is expected from the measurement errors, especially when considering that the literature sample is not homogeneously analyzed, and we have not taken into account any possible errors of the NLTE corrections, nor the variation in stellar ages.

The most metal-poor stars ([Fe/H] ≈−3\approx-3) are expected to have ages of ∼13\sim13 Gyr (e.g. Koutsouridou et al. [68]). Since the GSE merger is estimated to have happened around 10 Gyr ago (e.g. [50, 28]), this sets a lower limit of about ∼10\sim10 Gyr on the ages of the most metal-rich actinide-boost stars, which are known to belong to GSE (e.g. Ernandes et al. [26]). Recalling that log⁡ϵ(Th/Eu)\log\epsilon(\mathrm{Th/Eu}) ratio is a cosmo-chronometer7, the log⁡ϵ(Th/Eu)\log\epsilon(\mathrm{Th/Eu}) should change by an amount 0.108 dex if the age difference between stars is 5 Gyr, 0.065 dex in the case of 3 Gyr, and 0.043 dex in the case of 2 Gyr. Based on our measurements, we can thus put an upper limit of an age difference between the oldest and youngest actinide-boost stars in our sample of <3<3 Gyr, as this corresponds to the span of log⁡ϵ(Th/Eu)\log\epsilon(\mathrm{Th/Eu}) values in our sample. This is in agreement with our expectations of the stars belonging to the age range ∼\sim10-13 Gyr.

If we assume an average age of 11.5 ±\pm 1.0 Gyr for actinide-boost stars in our sample, the production value of the actinide-boost rr-process is log⁡ϵ(Th/Eu)=+0.046±0.021\log\epsilon(\mathrm{Th/Eu}) = +0.046 \pm0.021. With larger samples and robust measurements and NLTE calculations, there is large potential for accurate differential ages of different populations of actinide-boost stars.

Finally, we note that the actinide-boost stars identified so far are all older than ∼\sim10 Gyr irrespective of their metallicities. Observational scarcity of young actinide-boost stars is likely caused by their signature being diluted and washed out by the homogenization in more evolved galactic environments. In other words, their window of visibility is restricted to the highly structured, poorly mixed epoch of the early galactic assembly. However, systematic searches of actinide-boost stars among more metal-rich younger stars have yet to be carried out.

The actinide-boost source

Perhaps the most fundamental question is: What does this plateau tell us about the origin of the actinide-boost rr-process? In § 6, we explored one plausible scenario of the formation of such stars, but that does not necessarily mean that all other options can be excluded.

Firstly, the observational gap between actinide-boost and actinide-normal stars indicates that they are most likely formed from gas enriched by two separate sites. Alternatively, there would have to be a very sharp change in either physical parameters (such as YeY_{\mathrm{e}}) within a given site, or a sharp change in log⁡ϵ(Th/Eu)\log\epsilon(\mathrm{Th/Eu}) as a function of other parameters, but current models do not predict this (e.g. Eichler et al. 2019). These new results indicate that log⁡ϵ(Th)\log\epsilon(\mathrm{Th}) does not show a continuous change from actinide-normal to actinide-boost stars, as generally assumed from LTE data (e.g. Holmbeck et al. [55]).

Based on several indirect lines of evidence, here we favor a scenario where NSM produce the actinide-boost stars, and an alternative primary site associated with massive stars (prompt) is responsible for actinide-normal stars. The key arguments are as follows:

1. Yields: According to state-of-the-art models, NSMs are typically predicted to have higher Th/Eu ratios than other proposed rr-process sites, such as MRNe, collapsars, and magnetar giant flares (e.g. [126, 84, 56, 81, 95, 89] (Rigrigiano et al. 2024)).

2. Galactic Archaeology: Observations of the rr-process in the Milky Way and its satellite galaxies can only be explained by adopting at least two sources, NSM (delayed) and a prompt source associated with massive stars (e.g. [115, 117]). In the Milky Way, the dominating source should be prompt to explain the general evolution of Eu (e.g. [65, 82] (Côté et al. 2019)). Given the rarity of actinide-boost stars, it is likely that their source is associated with the less dominant site, i.e. NSM. If NSM are not the site of actinide-boost stars, we would require at minimum 3 different sources to explain the abundance patterns in stars: NSM, an actinide-normal prompt source (main contributor of the Milky Way evolution), and an actinide-boost prompt source.

3. Less efficient star formation: Most (if not all) of the known actinide-boost stars in the Milky Way are accreted from smaller galaxies. It is well known that delayed sources are particularly prominent in such systems, as seen by strong effects of SNIa and AGB stars, due to a combination of less efficient star formation and them being prone to lose their gas (e.g. Tolstoy et al. [123]; Skúladóttir & Salvadori [117]. Thus, this result points to a delayed source of the actinide-boosted rr-process, i.e. NSM.

Finally, it is worth noting that the extreme narrowness of the plateau can be used to constrain the physical conditions of any suggested actinide-boost site. However, the outlier J1424 (Placco et al. [91]) seems to indicate that although a fixed intrinsic value is the most common for actinide-boost events, it might not be universal. An outlier such as this could stem from, e.g. a black hole-neutron star merger (Wanajo et al. [125]), or a NSM with unusual conditions (e.g. in relation to the mass involved). But to draw firm conclusions about the different sources, a population study of actinide-boost stars is needed, especially in relation to their frequency in different formation environments and at different metallicity ranges.

Future prospects

To fully explore and understand the actinide-boost stars and which astrophysical sites are capable of reproducing the observations, we need to overcome both theoretical and observational limitations. Improved NLTE calculations for the heavy elements, covering a large range in stellar parameters, are needed to derive more reliable abundance ratios. Together with state-of-the-art 3D atmospheric models, this can test the robustness of the observed bimodality. In addition, accurate and robust theoretical yields are needed for all proposed rr-production sites.

On the observational side, the sample sizes are limited in all environments, especially in dwarf galaxy populations, where only a few measurements are available (e.g. Aoki et al. [2]; Hansen et al. [45]; Ji & Frebel 2018), and high-metallicity halo stars, where measurements become exceedingly challenging. The available data are expected to improve drastically with upcoming spectroscopic surveys like WEAVE (Dalton et al. [22]) and 4MOST - especially the 4MOST dedicated surveys for dwarf galaxies (Skúladóttir et al. [116]) and the high-resolution halo survey (Christlieb et al. [17]).

These surveys, combined with high-resolution, high-quality follow-up observations using blue sensitive spectrographs like VLT-CUBES (Evans et al. [27]; Ernandes et al. [24]), and the next generation of high-resolution instruments on the extremely large telescopes such as ELT-ANDES (Marconi et al. [76]), and in the 8m-class telescopes such as HRMOS (Magrini et al. [75]), will enable us to access the Th lines for fainter targets, significantly expanding the sample of stars.

Conclusions

From the historical sample of Nissen & Schuster [86], we have measured high-precision abundances of Th and Eu in 50 stars. This contains Th measurements for 23 stars in Gaia Sausage Enceladus (GSE), making it the largest sample of [Th/Eu] measurements in any external galaxy to date (§ 3). Among GSE stars, we identify three actinide-boost stars, which are the most metal-rich actinide-boost stars discovered to date ([Fe/H] >> −-1.6). Meanwhile, no actinide-boost star was among our in-situ Milky Way sample of 27 stars (§ 5.1). In fact, most (if not all) actinide-boost stars in our sample originate from outside the Milky Way (§ 4).

After applying NLTE corrections to our sample and stars from the literature, a bimodal distribution emerges: actinide-boost stars lie on a narrow plateau with log ε(Th/Eu) = −0.2, clearly separated from actinide-normal stars at log ε(Th/Eu) < −0.35. Out of 13 actinide-boost stars in our samples, 12 fall on this plateau, which has a constant value over a large range in metallicity, −3.5 ≲ [Fe/H] ≲ −1 (Sec. 5.2).

The narrowness of the plateau (σ = 0.02) allows us to estimate the age range of the sample as ≲ 3 Gyr. Adopting an average age of 11.5 ± 1.0 Gyr, we estimate the intrinsic value of the actinide-boosted rr-process as log ε(Th/Eu) = +0.046 ± 0.021 (§ 7.1). This provides very strong constraints on theoretical yield predictions for any rr-process site with boosted actinide production. Our results indicate that log ε(Th/Eu) is extremely promising as a cosmic clock for actinide-boost stars, which seem to have a robust production ratio. For actinide-normal stars, however, the distribution should be broadened by different mixing ratios of multiple rr-process events of the two types (actinide-boost, and actinide-normal).

Our chemical evolution models (§ 6.1) show that the formation of actinide-boost stars, at both low and high metallicity, can be explained in a scenario with two rr-process sources: 1) Neutron star mergers (NSM), which are delayed and have enhanced [Th/Eu] ratios; 2) a prompt source with lower [Th/Eu] ratios. In this scenario, a flat plateau arises whenever a single NSM dominates the rr-process content of the gas, either in a small system or in a gas-depleted region of a larger galaxy such as GSE. The absolute level of the plateau is instead set by the log ε(Th/Eu) production ratio of the actinide-boost rr-process source, which currently remains theoretically uncertain and can potentially be better constrained by our observations. Our model is not sensitive to which prompt site is adopted, and we draw no conclusions on which is the most likely prompt source out of those proposed in the literature. For the production site of the actinide-boost stars, here we favour NSMs, as these are predicted to produce high actinides, and this is also supported by arguments based on Galactic chemical evolution (see § 7.2).

Dedicated population studies of large samples are still needed to better understand the origin of actinide-boost stars (§ 7.3). Here we have demonstrated that accurate NLTE abundances are essential to provide the full picture, as LTE abundances give a very skewed view - overestimating their fraction and hiding the bimodality of [Th/Eu] abundances. Finally, our results show that even though our understanding has improved significantly in recent years, the rr-process still offers unexpected surprises.

Acknowledgements. We thank the anonymous referee for the feedback. This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No. 101117455). H.E. were supported by a project grant from the Knut and Alice Wallenberg Foundation (KAW 2020.0061 Gaclick Time Machine, PI Feltzing). H.E. is supported by foundations managed by Kungliga Vetenskapsakademien (reference AST2024-0042) and supported by The Kungl. Fysiografiska Sällskapet i Lund (reference 20240013). This project was supported by funds from the Crafoord Foundation (reference 20230890). AA, MM and JK acknowledge support by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – Project-ID 279384907 – SFB 1245. Based on data collected using VLT/UVES under the programs: PI: Á. Skúladóttir, 109.22 and 110.240 with the additional data from ESO Science Archive Facility from 8.B-0475, 076.B-0133, 090.B-0605, 109.23, 71.B0529, and 68.D-0094.

Stellar parameters, abundances, and kinematic parameters

Stellar atmospheric parameters, elemental abundances, and kinematic parameters of the primary sample and the literature sample.

ID NS24TeffT_{\mathrm{eff}} (K)log⁡g\log g (dex)vmicv_{\mathrm{mic}} (km s−1^{-1})[Fe/H] (dex)[Mg/Fe]†^{\dagger} (dex)log⁡ϵ(Eu)LTE\log\epsilon(\mathrm{Eu})_{\mathrm{LTE}} (dex)log⁡ϵ(Th)LTE\log\epsilon(\mathrm{Th})_{\mathrm{LTE}} (dex)log⁡ϵ(Th/Eu)LTE\log\epsilon(\mathrm{Th/Eu})_{\mathrm{LTE}} (dex)ΔNLTE(Eu)\Delta_{\mathrm{NLTE}}(\mathrm{Eu}) (dex)log⁡ϵ(Th/Eu)NLTE∗\log\epsilon(\mathrm{Th/Eu})_{\mathrm{NLTE}}^{*} (dex)Pop
G 112−4362094.021.17−1.270.34−0.46−0.450.010.19−0.17GSE
G 13−3852634.540.90−0.880.35−0.09−0.40−0.320.14−0.45In situ halo
G 66−2252974.460.78−0.880.020.01−0.43−0.440.10−0.53GSE
G 57−0757554.330.99−0.480.290.23−0.13−0.360.15−0.50In situ halo
G 18−2854434.490.88−0.850.31−0.07−0.40−0.330.11−0.43In situ halo
G 170−5661124.111.39−0.940.23−0.10−0.42−0.330.17−0.49GSE
G 159−5057134.441.03−0.940.36−0.14−0.57−0.430.15−0.57In situ halo
G 31−5557314.351.26−1.120.28−0.40−0.68−0.280.16−0.43In situ halo
G 82−0553384.510.80−0.780.000.16−0.30−0.460.10−0.55GSE
G 20−1561624.321.50−1.500.27−0.37−0.90−0.530.11−0.63GSE
CD−33 333761123.861.56−1.370.44−0.50−0.87−0.370.12−0.48TD
CD−43 681060594.321.24−0.440.350.28−0.01−0.290.10−0.38In situ halo
CD−57 163359814.291.08−0.910.06−0.08−0.45−0.380.10−0.47GSE
G 24−2558283.861.20−1.400.34−0.26−0.55−0.290.13−0.41GSE
G 63−2661754.171.65−1.580.44−0.64−1.30−0.660.16−0.81In situ halo
G 75−3161354.021.28−1.040.24−0.15−0.50−0.350.19−0.53GSE
HD 10372360504.201.11−0.810.170.03−0.30−0.330.11−0.43GSE
HD 10500458524.351.09−0.830.16−0.07−0.50−0.430.15−0.57GSE
HD 10651663274.431.18−0.690.460.09−0.30−0.390.10−0.48TD
HD 11198058783.981.39−1.090.43−0.31−0.69−0.380.10−0.47In situ halo
HD 11367957614.051.37−0.660.400.11−0.33−0.440.13−0.56In situ halo
HD 12055954864.581.05−0.910.30−0.09−0.45−0.360.11−0.46TD
HD 12100457554.431.16−0.710.360.14−0.35−0.490.10−0.58In situ halo
HD 13247557503.771.37−1.510.46−0.675−1.05−0.380.17−0.54In situ halo
HD 14881659234.171.33−0.740.380.03−0.30−0.330.14−0.46In situ halo
HD 15948258294.371.21−0.740.330.05−0.35−0.40.13−0.52In situ halo
HD 16381055924.611.17−1.220.11−0.225−0.51−0.290.15−0.43GSE
HD 17709553494.390.90−0.740.380.02−0.50−0.520.13−0.64In situ halo
HD 17962659254.141.49−1.060.41−0.32−0.59−0.270.16−0.42In situ halo
HD 18955857073.831.29−1.140.41−0.33−0.67−0.340.11−0.44TD
HD 19390157294.431.31−1.110.10−0.18−0.62−0.440.12−0.55GSE
HD 19459860184.341.40−1.110.21−0.25−0.67−0.420.14−0.55GSE
HD 19928959154.301.21−1.050.39−0.28−0.59−0.320.11−0.42TD
HD 20565057934.351.17−1.190.36−0.36−0.82−0.460.12−0.57TD
HD 21961759834.281.42−1.460.29−0.56−0.95−0.390.11−0.49In situ halo
HD 2287958594.291.20−0.860.36−0.05−0.45−0.410.10−0.50TD
HD 28424862714.211.51−1.590.31−0.65−0.67−0.020.18−0.19GSE
CD−51 462862664.291.31−1.320.22−0.30−0.75−0.460.17−0.62GSE
HD 356761804.011.40−1.170.25−0.07−0.42−0.350.18−0.52GSE
HD 5175458574.351.30−0.580.350.20−0.12−0.320.10−0.41In situ halo
HD 5939261373.881.73−1.620.40−0.47−1.00−0.530.18−0.70GSE
HD 7693259774.171.30−0.870.420.02−0.55−0.570.10−0.66TD
G 16−2056253.641.50−1.420.22−0.48−0.82−0.340.13−0.46GSE
CD−61 28258694.341.19−1.250.21−0.27−0.77−0.500.12−0.61GSE
Archival data
G 87−1362174.111.42−1.110.23−0.10−0.17−0.070.15−0.21GSE
G 98−5359544.261.20−0.890.25−0.03−0.35−0.320.19−0.50GSE
G 161−7361083.991.26−1.010.20−0.20−0.46−0.260.17−0.42GSE
G 53−4159564.341.35−1.210.22−0.25−0.52−0.270.18−0.44GSE
G 188−2261164.201.42−1.330.48−0.54−0.72−0.180.20−0.37In situ halo
HD 2570459744.301.30−0.860.36−0.02−0.45−0.430.15−0.57TD

Table A.1. Stellar atmospheric parameters and elemental abundances of the stellar sample analyzed in this work.

Note: †^{\dagger} LTE values, NLTE corrections are negligible for the parameter range of the sample. ∗^{*} ΔNLTE(Th)∼0.01\Delta_{\mathrm{NLTE}}(\mathrm{Th}) \sim0.01 dex for the parameter range of the sample.

IDTeffT_{\mathrm{eff}} (K)log gg (dex)[Fe/H] (dex)[Mg/Fe]NLTE^{\mathrm{NLTE}} (dex)log ϵ(Eu)LTE\epsilon(\mathrm{Eu})_{\mathrm{LTE}} (dex)log ϵ(Th)LTE\epsilon(\mathrm{Th})_{\mathrm{LTE}} (dex)log ϵ(Th/Eu)LTE\epsilon(\mathrm{Th/Eu})_{\mathrm{LTE}} (dex)ΔNLTE(Eu)\Delta_{\mathrm{NLTE}}(\mathrm{Eu}) (dex)ΔNLTE(Th)\Delta_{\mathrm{NLTE}}(\mathrm{Th}) (dex)log ϵ(Th/Eu)NLTE\epsilon(\mathrm{Th/Eu})_{\mathrm{NLTE}} (dex)Ref.
HE 2327−564250502.34−2.780.14−1.29−1.67−0.380.160.05−0.491
2MASS J09544277+524641444100.61−2.960.69−1.16−1.79−0.630.140.09−0.682
HE 2252−422547101.65−2.630.20−1.3−1.63−0.330.140.08−0.393
BPS CS 29491−06953002.80−2.510.37−1.03−1.43−0.400.110.05−0.464
HE 1219−031250602.30−2.970.50−0.98−1.14−0.160.110.06−0.215
BPS CS 30315−02945700.99−3.400.71−2.24−2.45−0.210.100.12−0.196
LAMOST J080452.84+574019.645851.41−2.380.42−1.06−1.28−0.220.060.07−0.217
BPS CS 22953−00348601.65−3.000.42−1.69−1.92−0.230.060.09−0.208
BD−15 578145500.70−2.870.54−2.28−2.73−0.450.270.16−0.568
BD+10 249548901.85−2.140.41−1.34−1.77−0.430.120.08−0.478
HD 2629744001.10−1.720.23−1.30−1.73−0.430.120.11−0.448
HD 4528252302.90−1.730.47−0.70−1.08−0.380.100.04−0.448
HD 17530549202.30−1.560.30−0.64−1.02−0.380.100.06−0.428
2MASS J14244532−254247147621.58−3.390.83−1.25−1.210.040.100.090.039
BPS CS 29497−00450902.40−2.640.38−0.45−0.96−0.510.110.05−0.5710
BD+17 324852001.80−2.000.49−0.67−1.18−0.510.110.08−0.5411
BPS CS 22892−05248001.50−3.100.44−0.95−1.57−0.620.110.09−0.6412
HE 1523−090146301.00−2.950.63−0.62−1.20−0.580.110.13−0.5613
HD 22117045101.00−2.190.46−0.86−1.46−0.600.100.12−0.5814
BPS CS 31082−00148251.50−2.900.50−0.76−0.98−0.220.060.07−0.2115
HD 10857749001.10−2.200.40−1.49−2.00−0.510.160.16−0.5116
HD 11544445000.70−3.000.62−1.82−2.37−0.550.230.15−0.6316
BD+08 285645500.70−2.000.59−1.15−1.65−0.500.110.15−0.4616
HD 626846001.00−2.630.57−1.56−1.93−0.370.190.13−0.4317
HD 11018442400.30−2.520.23−1.91−2.50−0.590.220.20−0.6117
HD 18647847201.60−2.50.50−1.34−1.85−0.510.130.08−0.5617
BPS CS 30306−13251102.50−2.420.43−1.02−1.12−0.100.100.04−0.1617
BPS CS 31078−01852572.75−2.850.55−1.17−1.35−0.180.100.05−0.2318
RAVE J005327.8−02531743700.56−2.160.51−1.25−2.14−0.890.130.17−0.8519
RAVE J024620.1−15184249481.93−2.450.26−0.64−1.40−0.760.110.08−0.7919
RAVE J031337.3−10205647571.47−2.050.42−1.11−1.30−0.190.110.09−0.2119
RAVE J034346.3−09241743800.46−1.920.52−1.02−1.52−0.500.110.18−0.4319
RAVE J141015.9−03435549312.00−2.060.40−0.87−1.59−0.720.100.07−0.7519
2MASS J15211026−060756647071.33−2.000.39−0.55−1.36−0.810.070.12−0.7619
RAVE J153830.9−18042447521.63−2.090.43−0.30−0.95−0.650.060.08−0.6319
2MASS J22562536−071956245581.20−2.260.48−0.64−1.66−1.020.080.10−1.0019
HD 2052201.89−1.780.60−0.35−0.85−0.500.100.07−0.5320
SMSS J200322.54−114203.351752.44−3.500.32−1.21−1.31−0.100.120.05−0.1721
HD 22292553662.54−1.470.440.38−0.06−0.440.080.06−0.4622
2MASS J22132050−513738555092.28−2.200.720.770.22−0.550.110.05−0.6123
BD+29 235647601.30−1.550.38−0.69−1.07−0.380.110.12−0.3724
BD+30 261143300.60−1.450.28−0.36−0.98−0.620.090.16−0.5524
CD−68 188147801.50−1.630.41−0.63−1.19−0.560.100.09−0.5725
HE 0050−091853102.00−1.370.37−0.14−0.71−0.570.130.11−0.5925
HE 0324−012251452.45−2.410.51−0.77−1.32−0.550.100.05−0.6025
CD−37 206648751.75−2.020.51−1.17−1.75−0.580.110.09−0.6025
CD−22 769645800.95−2.480.35−1.69−2.26−0.570.190.12−0.6425
HE 0007−175252152.65−2.36−0.19−0.6−1.16−0.560.100.04−0.6225
TYC 3431−689−148501.40−2.050.51−0.73−1.34−0.610.090.09−0.6126
TYC 1191−918−145001.20−2.150.71−0.83−1.24−0.410.090.09−0.4126
LAMOST J112456.61+453531.351802.70−1.27−0.200.38−0.19−0.570.070.05−0.5927
BPS CS 29529−008950142.16−2.460.43−0.15−0.55−0.400.100.06−0.4428
HD 1083751002.67−2.370.45−1.37−1.99−0.620.110.04−0.6929
2MASS J14325334−412549449001.70−2.970.66−1.01−1.49−0.480.110.09−0.5030
2MASS J20050670−305744544300.60−3.030.58−1.57−2.18−0.610.190.17−0.6330
2MASS J15213995−353809458502.10−2.800.55−0.04−0.60−0.560.180.14−0.6031
2MASS J05383296−590428058192.09−2.550.66−0.78−1.16−0.380.160.14−0.4032
RAVE J094921.8−16172248551.60−2.220.54−1.09−1.70−0.610.110.09−0.6333
LAMOST J020623.21+494127.941301.52−0.540.171.180.48−0.700.120.04−0.7834
RAVE J20843.2−00233346301.20−2.910.48−0.75−1.24−0.490.100.11−0.4835

Table A.2. Data for the literature sample.

Note: †The NLTE corrections are calculated using the NL iTE tool by adopting the grid of Lind et al. [72].

References. 1. Mashonkina et al. [77], 2. Shah et al. [109], 3. Mashonkina et al. [78], 4. Hayek et al. [49], 5. Roederer et al. [100], 6. Siqueira Mello et al. [113], 7. Lin et al. [71], 8. Roederer et al. [102], 9. Placco et al. [91], 10. Christlieb et al. [18], 11. Cowan et al. [20], 12. Sneden et al. [119], 13. Frebel et al. [32], 14. Ivans et al. [59], 15. Hill et al. [51], 16. Johnson & Bolte [62], 17. Honda et al. [58], 18. Lai et al. [69], 19. Sakari et al. [106], 20. Hanke et al. [44], 21. Yong et al. [130], 22. Roederer et al. [103], 23. Roederer et al. [98], 24. Roederer et al. [104], 25. Gull et al. [42], 26. Saraf et al. [107], 27. Xing et al. [129], 28. da Silva & Smljanic [21], 29. Roederer et al. [101], 30. Cain et al. [12], 31. Cain et al. [13], 32. Hansen et al. [46], 33. Gull et al. [41], 34. Xie et al. [128], 35. Placco et al. [92].

Statistical Analysis

To corroborate the robustness of the bimodality and quantify the populations, we fitted the log⁡ϵ(Th/Eu)\log\epsilon(\mathrm{Th}/\mathrm{Eu}) distributions (both LTE and NLTE) with Gaussian Mixture Models (GMM), followed by a Bayesian posterior probability distribution analysis.

First, we performed a GMM analysis for K=1K = 1, 2, and 3 component models and the Bayesian Information Criterion (BIC) was evaluated to determine the optimal number of populations. In NLTE, the two-component model is decisively preferred over both the one- and three-component models. The two-component model minimizes the BIC value with a ΔBIC=BIC1−BIC2=22.38\Delta\mathrm{BIC} = \mathrm{BIC}_{1} - \mathrm{BIC}_{2} = 22.38, corresponding to a Bayes Factor ln⁡(B21)=11.20\ln(B_{21}) = 11.20. A ln⁡(B21)>5\ln(B_{21}) > 5 (ΔBIC>10\Delta\mathrm{BIC} > 10) provides decisive statistical evidence in favor of two distinct populations over a single continuous population ([64]). The LTE distribution is strongly favored by a unimodal distribution with a ΔBIC=−14.45\Delta\mathrm{BIC} = -14.45 and Bayes factor ln⁡(B12)=7.23\ln(B_{12}) = 7.23.8 The 2-component GMM identified two well-separated Gaussian distributions in NLTE with mean (μ\mu) and standard deviations (σ\sigma): −0.54±0.11-0.54 \pm0.11 (standard peak; μ1±σ1\mu_{1} \pm\sigma_{1}) comprising 89.1% of the sample and −0.20±0.02-0.20 \pm0.02 (actinide-boost peak; μ2±σ2\mu_{2} \pm\sigma_{2}) comprising 10.9% of the sample. The LTE distribution yielded a single Gaussian with μ±σ=−0.43±0.18\mu\pm\sigma= -0.43 \pm0.18.

The standard deviations obtained from the GMM are the observed widths of the log⁡ϵ(Th/Eu)\log\epsilon(\mathrm{Th}/\mathrm{Eu}) distributions. To deconvolve the observational uncertainties (Δ∼0.07\Delta\sim0.07 dex) from this dispersion and to obtain the intrinsic (cosmic) scatter of the log⁡ϵ(Th/Eu)\log\epsilon(\mathrm{Th}/\mathrm{Eu}) distribution, we performed a Bayesian Markov Chain Monte Carlo (MCMC) analysis using the parameters obtained from GMM as priors. The resulting posterior probability distributions for the intrinsic parameters, μ\mu and σint\sigma_{\mathrm{int}}, are shown in Fig. B.1. In NLTE, intrinsic parameters, along with the 1σ1\sigma confidence level, obtained from the posterior distribution are μ1=−0.54±0.01\mu_{1} = -0.54 \pm0.01 with σint,1=0.08±0.01\sigma_{\mathrm{int},1} = 0.08 \pm0.01 for the standard peak and μ2=−0.20±0.02\mu_{2} = -0.20 \pm0.02 with σint,2=0.00±0.02\sigma_{\mathrm{int},2} = 0.00 \pm0.02 for the actinide-boost peak. The probability density in the observational gap −0.35<log⁡ϵ(Th/Eu)NLTE<−0.25-0.35 < \log\epsilon(\mathrm{Th}/\mathrm{Eu})_{\mathrm{NLTE}} < -0.25 is zero at 94.6% (1.9σ1.9\sigma) confidence level, as inferred from a posterior predictive distribution. The LTE distribution is represented by the intrinsic parameters μ=−0.43±0.02\mu= -0.43 \pm0.02 and σint=0.16±0.01\sigma_{\mathrm{int}} = 0.16 \pm0.01. In all cases, the μ\mu values are identical to those obtained from GMM despite the slightly lower values (by ∼0.02\sim0.02 dex) obtained for σint\sigma_{\mathrm{int}}.

Figure B.1. Posterior probability distributions of μ\mu and σint\sigma_{\mathrm{int}} of log⁡ϵ(Th/Eu)\log\epsilon(\mathrm{Th}/\mathrm{Eu}) distribution based on Bayesian analysis for LTE abundances, standard, and actinide-boost peaks (from top to bottom).

IDGaia DR3 IDEnergy (km2^2 s−2^{-2})LzL_z (km s−1^{-1} kpc)Eccentricityzmax⁡z_{\max} (kpc)ULSRU_{\mathrm{LSR}} (km s−1^{-1})VLSRV_{\mathrm{LSR}} (km s−1^{-1})WLSRW_{\mathrm{LSR}} (km s−1^{-1})VspaV_{\mathrm{spa}} (km s−1^{-1})
HE 2327-56426495850379767072128-26994.70-699.000.6614.65-65.32-354.25-235.14430.18
2MASS J09544277+5246414828438619475671936-23790.37-1710.380.409.89-146.20-385.83-109.34426.84
HE 2252-422565429097866724623367169.84262.350.8150.20-116.23-162.99323.73380.63
BPS CS 29491-06966009713192431741443655.79-1440.490.7336.11-167.99-422.83276.14532.22
HE 1219-03123693297117190538368––0.127.54––––
BPS CS 30315-0292332019357414106624-22074.91-195.890.8918.36143.10-287.39175.00365.64
LAMOST J080452.84+574019.61081668723016932736-16409.2043.280.9911.54245.79-237.30-104.85357.38
BPS CS 22953-0034710594687144052096-24918.04-748.280.7712.17278.02-249.803.43373.77
BD-15 57816887351938881045504-70825.33355.430.761.01-3.72-167.33-25.39169.29
BD+10 24953734341989334114304––0.081.28––––
HD 262975097726306014470016-51048.371119.670.382.71-34.16-87.8881.59124.69
HD 452823125113282265023488-35306.81438.730.891.05-238.06-167.54-34.89293.18
HD 1753052268048503896398720-14917.681179.850.4818.56-53.41-72.34-278.96293.10
2MASS J14244532-25424716271613367058424064-46579.45745.910.247.04-83.96-18.8256.14102.74
BPS CS 29497-0042322729725405593728-30973.841170.190.617.58-138.86-64.28-111.44189.30
BD+17 32484553184509407224576-66644.32-29.250.981.62-39.67-221.4863.26233.73
BPS CS 22892-0526826025986350385280-27857.93-167.490.9314.68-169.96-186.18-212.62329.78
HE 1523-09016317828550897175936-33972.75-1278.190.447.62-130.40-443.88-134.26481.72
HD 2211702869759781250083200-51668.09825.960.650.92134.64-125.77-33.27187.23
BPS CS 31082-0012451773941958712192-22842.87-1288.420.3913.74125.51-374.03-191.06438.36
HD 1085773907872613852996864-31510.47-438.700.6514.14-181.05-280.57-177.54378.18
HD 1154441474455748663044736-50203.79430.820.831.55166.65-169.4316.20238.20
BD+08 28561175731289937614080-40369.551066.610.622.69184.92-69.36-17.41198.26
HD 62685033960575837265920-35195.921562.190.481.95147.53-24.87-35.18153.69
HD 1101843710786979933623424-55578.49611.970.545.09-25.29-143.01112.97183.99
HD 1864786871465473332110464-51469.25-660.470.721.72163.41-318.58-56.00362.39
BPS CS 30306-1321163455203958063872-52110.6788.000.573.49-132.23-70.18-2.61149.48
BPS CS 31078-018718978?-49125.451014.350.501.36-85.08-105.85-32.65139.60
RAVE J052728.0-025317259?-45613.64-1137.300.564.90-27.54-348.8294.64362.48
RAVE J115184.2-151842515?-34291.81-1168.620.7911.67-220.75-353.65-136.65437.38
RAVE J01337.3-102056516?-34291.81-588.620.844.22-184.58-280.08-73.15350.78
RAVE J034346.3-092417516?-24901.73620.480.866.60-209.32-147.47-88.47266.68
RAVE J?460?-42796.86714.400.665.21-129.34-333.33-222.20379.56
2MASS J1121026-0607566632?-62652.43-492.780.513.1660.31-287.5095.04308.75
RAVE J1538.0-180424625?-56216.61499.860.670.33108.77-176.647.42211.34
HE 214?210?-??1.000.98-223.83-217.77-239.94501.61
RAVE J?410?-21463.71-1190.350.5514.6696.89-429.7767.92376.18
SMS J200322.54-114203.3648?-33367.02-407.750.904.04-244.11-278.05-81.28376.40
HD 229025651?-28232.94237.590.813.72338.88-183.35-188.28371.49
2MASS J22132050-5137385164?-29678.94-463.570.7713.95156.20-281.36-185.88292.78
BD+92 336125?-13487.361202.920.5320.6110.70-60.23-286.32292.78
BD+30 261581?-11503.92820.580.6024.24-47.11-103.35-331.00349.94
CD-68 1881

Table A.3. Kinematics of the combined sample

IDGaia DR3 IDEnergy (km² s⁻²)LzL_z (km s⁻¹ kpc)Eccentricityzmax⁡z_{\max} (kpc)ULSRU_{\mathrm{LSR}} (km s⁻¹)VLSRV_{\mathrm{LSR}} (km s⁻¹)WLSRW_{\mathrm{LSR}} (km s⁻¹)VspaV_{\mathrm{spa}} (km s⁻¹)
HE 0050-09182474070010546861056-10250.79782.070.8618.24262.40-139.50171.07342.90
HE 0324-01223267948604442696448-15679.571221.280.4218.4618.37-80.54-268.13280.56
CD-37 20664820462183959704448-24397.31-43.730.4316.5384.81-218.66267.92356.07
CD-22 76965660878702718068864-47950.66-66.130.549.17-9.02-230.15-168.23285.22
HE 0007-17522366585258508216320-38011.53479.730.068.4912.97-160.02-228.15278.98
TYC 3431-689-11019002298308911360-41165.93-777.260.328.50-46.88-298.09-141.51333.29
TYC 1191-918-12795305423661094016-51899.61490.400.801.63130.14-175.7044.28223.08
LAMOST J112456.61+453531.3785388513000524160-32286.56-103.480.4614.0534.07-232.75121.53264.77
BPS CS 29529-00894678520146255821952-50862.60971.760.463.32-100.13-130.2633.91167.77
HD 108372467368555894065664––0.081.14––––
2MASS J14325334-412549461033741336615964162587.65858.140.9215.71-357.03-131.38-107.48395.33
2MASS J20050670-30574456749102714060650880-37429.04-1183.140.603.54-216.16-418.406.63470.98
2MASS J15213995-35380946199959495651403776-9042.921944.830.6413.9873.4781.37188.14217.75
2MASS J05383296-59042804759914999195512064-51333.23750.070.545.12-75.78-135.87-122.34197.91
RAVE J094921.8-1617225684624825278191744-38665.58-1261.400.523.90-74.35-384.5981.56400.11
LAMOST J020632.21+494127.9358364512036089472-31644.952183.740.040.6631.32-11.049.5434.55
RAVE J203843.2-0023334226556850750887168-14974.23-1432.300.6515.67-26.12-535.77-136.82553.58
BD+00 2058A3085891537839267328-16298.491022.680.6118.40137.48-92.50-252.47301.99
BD+02 25413700341138433848832-53019.50652.760.693.19109.08-137.8792.34198.58
BD+06 29321159108770069883136-25629.1355.170.991.08-284.14-213.1230.65356.51
BD+11 23693916257626962267136-61025.68372.560.811.38-73.16-174.05-54.56196.52
BD+11 47252728314787225296000-66205.54-108.680.940.09-21.67-233.516.35234.60
BD+18 34234550762289589343488-63193.17-397.490.790.70-60.80-269.66-36.43278.82
BD-01 3062494759795723713408-44944.79262.140.912.23-192.64-187.1864.48276.23
BD-03 562540830947835865728-64392.56332.720.820.85-19.40-178.42-41.84184.29
BD-06 8553202470247468181632-5168.78385.200.962.74-342.79-173.2139.35…
BD-08 45014165370622239900160-22706.041334.410.9611.31158.79-51.65-188.57251.78
BD+33 33875589912918539906048-40544.931441.360.144.2927.93-406.68-135.29144.01
CD-33 18105486868153073144576-52136.96111.320.802.55-102.43-166.68-81.28211.85
CD-57 16334228716122169160960-41471.29-629.680.981.23-297.72-240.68-31.94383.76
G 24-253298511540427371392-52991.14-1027.780.802.63-200.72-295.96-74.54365.29
G 62-263925628980417321984-30302.15185.150.566.25-243.43-349.4244.55364.17
G 75-31349413542922554016-62409.90179.900.911.33-69.51-198.36101.80328.90
HD 1037233486536729252158976-62901.13-66.130.962.36-34.73-228.6455.33244.23
HD 1050043510282289561616128-3000.5028.350.332.2364.50-185.69-78.53244.46
HD 1056165102948282902616288-24559.27264.770.956.22270.59-185.27-53.79343.10
HD 119806141991322784963072-61185.97-451.770.780.09-88.77-278.15-100.62343.10
HD 1136795871765483114779904-51143.201350.190.260.75-24.33-50.892.34291.98
HD 205596548341834824776064-52296.04-225.220.874.1563.33-251.43-40.7969.61
HD 1234756232043687720079616-54174.56-1093.870.411.4253.11-357.8361.13366.91
HD 148816433711091045299090-58095.90-323.100.842.10102.11-261.0072.16289.40
HD 1594824485972467412021376-34499.591343.050.564.10159.12-50.74101.11195.24
IDGaia DR3 IDEnergy (km2^2 s−2^{-2})LzL_z (km s−1^{-1} kpc)Eccentricityzmax⁡z_{\max} (kpc)ULSRU_{\mathrm{LSR}} (km s−1^{-1})VLSRV_{\mathrm{LSR}} (km s−1^{-1})WLSRW_{\mathrm{LSR}} (km s−1^{-1})VspaV_{\mathrm{spa}} (km s−1^{-1})
HD 1638104150413750729448320−31961.85108.50–1.57262.28−206.9535.68336.00
HD 1770954082356695359119616−53342.90574.940.760.43146.64−147.94−22.91209.56
HD 1796264263841431819084032−60101.04−72.800.971.12106.64−229.9849.14258.22
HD 1895584189078626829509248−56022.17896.660.551.0281.89−107.5347.85143.38
HD 1939016859076107589173120−52306.18−252.080.892.69−145.22−251.41−80.50301.29
HD 1945981752459532807704704−63394.91−332.140.830.32−65.64−261.38−20.09270.24
HD 1992896481426917515214464−52259.271329.260.280.24−26.53−53.11−16.3361.57
HD 2056506810862213470371072−49936.341097.120.530.16−119.23−81.979.09144.97
HD 2196174263841431819084032−14959.22−242.842.65317.85−251.43−41.18407.36407.36
HD 228793250489115709122560−50675.791134.420.480.69−99.82−78.61−35.08131.81
HD 28424852624073910834176−1544.74510.000.949.14−345.87−156.49−68.04385.68
HD 2989865358179685572496384−17163.32645.810.8813.97254.49−134.90165.23332.06
HD 35672427069874188580480−51824.90−234.070.911.28162.04−249.95−46.62301.50
HD 517543112836479029139200−36030.78570.200.850.33234.76−148.0613.50277.88
HD 593925586241315104190848−51050.07−796.980.680.45143.58−316.46−23.13348.27
HD 769325730371484020807808––––––––
Wolf 6154412702764879270656−33605.57−268.940.942.46253.16−255.9055.31364.19
CD−61 2824715919175280799616−23789.52−480.360.910.17287.56−278.31−0.45400.19
G 87−13940306024863622400−44257.05−399.840.870.42−196.37−268.6818.69333.31
G 98−533440045641893430912−50096.18−191.950.923.26−149.68−243.45−85.92298.42
G 161−733822808140853060224−38691.70−243.460.931.73227.81−245.5646.54338.18
G 53−413855475490307708032−59975.66−227.690.873.95−25.68−248.66−103.69270.63
G 188−221800518563283501440−42409.931003.060.652.05170.71−96.9065.99207.09
HD 257044682867168555426944−45565.051333.020.450.05−118.40−53.99−3.06130.16

Table A.3 (continued).

References

  1. [1]Anders, E. & Grevesse, N. 1989, Geochim. Cosmochim. Acta, 53, 197
  2. [2]Aoki, W., Honda, S., Sadakane, K., & Arimoto, N. 2007, PASJ, 59, L15
  3. [3]Asplund, M., Amarsi, A. M., & Grevesse, N. 2021, A&A, 653, A141
  4. [4]Astropy Collaboration, Price-Whelan, A. M., Lim, P. L., et al. 2022, ApJ, 935, 167
  5. [5]Astropy Collaboration, Price-Whelan, A. M., Sipőcz, B. M., et al. 2018, AJ, 156, 123
  6. [6]Astropy Collaboration, Robitaille, T. P., Tollerud, E. J., et al. 2013, A&A, 558, A33
  7. [7]Beers, T. C. & Christlieb, N. 2005, ARA&A, 43, 531
  8. [8]Bovy, J. 2015, ApJS, 216, 29
  9. [9]Bradamante, F., Matteucci, F., & D’Ercole, A. 1998, A&A, 337, 338
  10. [10]Burbidge, E. M., Burbidge, G. R., Fowler, W. A., & Hoyle, F. 1957, Reviews of Modern Physics, 29, 547
  11. [11]Burris, D. L., Pilachowski, C. A., Armandroff, T. E., et al. 2000, ApJ, 544, 302
  12. [12]Cain, M., Frebel, A., Gull, M., et al. 2018, ApJ, 864, 43
  13. [13]Cain, M., Frebel, A., Ji, A. P., et al. 2020, ApJ, 898, 40
  14. [14]Castelli, F. & Kurucz, R. L. 2003, in Proceedings of the IAU Symp. No 210, Vol. 210, Modelling of Stellar Atmospheres, ed. N. Piskunov, W. W. Weiss, & D. F. Gray, A20
  15. [15]Cehula, J., Thompson, T. A., & Metzger, B. D. 2024, MNRAS, 528, 5323
  16. [16]Cescutti, G., Romano, D., Matteucci, F., Chiappini, C., & Hirschi, R. 2015, A&A, 577, A139
  17. [17]Christlieb, N., Battistini, C., Bonifacio, P., et al. 2019, arXiv preprint arXiv:1903.02468DOI
  18. [18]Christlieb, N., Beers, T. C., Barklem, P. S., et al. 2004, A&A, 428, 1027
  19. [20]Cowan, J. J., Sneden, C., Burles, S., et al. 2002, ApJ, 572, 861
  20. [21]da Silva, A. R. & Smiljanic, R. 2025, A&A, 696, A122
  21. [22]Dalton, G., Trager, S., Abrams, D. C., et al. 2016, in Ground-based and Airborne Instrumentation for Astronomy VI, Vol. 9908, SPIE, 394–403
  22. [24]Ernandes, H., Evans, C., Barbuy, B., et al. 2020, in Ground-based and Airborne Instrumentation for Astronomy VIII, Vol. 11447, SPIE, 1327–1337
  23. [25]Ernandes, H., Feuillet, D., Feltzing, S., & Skúladóttir, Á. 2024, A&A, 691, A333
  24. [26]Ernandes, H., Skúladóttir, Á., Feltzing, S., & Feuillet, D. 2025, A&A, 703, A256
  25. [27]Evans, C. J., Barbuy, B., Castilho, B., et al. 2018, in Ground-based and Airborne Instrumentation for Astronomy VII, ed. C. J. Evans, L. Simard, & H. Takami, Vol. 10702, International Society for Optics and Photonics (SPIE), 107022EDOI
  26. [28]Feuillet, D. K., Sahlholdt, C. L., Feltzing, S., & Casagrande, L. 2021, MNRAS, 508, 1489
  27. [29]Fishlock, C. K., Yong, D., Karakas, A. I., et al. 2017, MNRAS, 466, 4672
  28. [30]François, P., Depagne, E., Hill, V., et al. 2007, A&A, 476, 935
  29. [31]François, P., Matteucci, F., Cayrel, R., et al. 2004, A&A, 421, 613
  30. [32]Frebel, A., Christlieb, N., Norris, J. E., et al. 2007, ApJ, 660, L117
  31. [33]Freudling, W., Romaniello, M., Bramich, D. M., et al. 2013, A&A, 559, A96
  32. [34]Fujibayashi, S., Sekiguchi, Y., Shibata, M., & Wanajo, S. 2023, ApJ, 956, 100
  33. [35]Gaia Collaboration, Prusti, T., de Bruijne, J. H. J., et al. 2016, A&A, 595, A1
  34. [36]Gaia Collaboration, Vallenari, A., Brown, A. G. A., et al. 2023, A&A, 674, A1
  35. [37]Gehrels, N. 1986, ApJ, 303, 336
  36. [38]Ginsburg, A., Sipőcz, B. M., Brasseur, C. E., et al. 2019, AJ, 157, 138
  37. [39]Goriely, S., Bauswein, A., & Janka, H.-T. 2011, ApJ, 738, L32
  38. [40]Gudin, D., Shank, D., Beers, T. C., et al. 2021, ApJ, 908, 79
  39. [41]Gull, M., Frebel, A., Cain, M. G., et al. 2018, ApJ, 862, 174
  40. [42]Gull, M., Frebel, A., Hinojosa, K., et al. 2021, ApJ, 912, 52
  41. [43]Guo, Y., Storm, N., Bergemann, M., et al. 2025, A&A, 693, A211
  42. [44]Hanke, M., Hansen, C. J., Ludwig, H.-G., et al. 2020, A&A, 635, A104
  43. [45]Hansen, C. J., El-Souri, M., Monaco, L., et al. 2018, ApJ, 855, 83
  44. [46]Hansen, T. T., Roederer, I. U., Shah, S. P., et al. 2025, A&A, 697, A127
  45. [47]Hansen, T. T., Simon, J. D., Marshall, J. L., et al. 2017, ApJ, 838, 44
  46. [48]Hattori, K., Okuno, A., & Roederer, I. U. 2023, ApJ, 946, 48
  47. [49]Hayek, W., Wiesendahl, U., Christlieb, N., et al. 2009, A&A, 504, 511
  48. [50]Helmi, A., Babusiaux, C., Koppelman, H. H., et al. 2018, Nature, 563, 85
  49. [51]Hill, V., Plez, B., Cayrel, R., et al. 2002, A&A, 387, 560
  50. [52]Hirai, Y., Beers, T. C., Chiba, M., et al. 2022, MNRAS, 517, 4856
  51. [53]Hirai, Y., Beers, T. C., Lee, Y. S., et al. 2025, ApJ, 990, 125
  52. [54]Holmbeck, E. M., Beers, T. C., Roederer, I. U., et al. 2018, ApJ, 859, L24
  53. [55]Holmbeck, E. M., Frebel, A., McLaughlin, G. C., et al. 2019a, ApJ, 881, 5
  54. [56]Holmbeck, E. M., Sprouse, T. M., Mumpower, M. R., et al. 2019b, ApJ, 870, 23
  55. [57]Holmbeck, E. M., Surman, R., Frebel, A., et al. 2020, Journal of Physics: Conference Series, 1668, 0612020
  56. [58]Honda, S., Aoki, W., Kajino, T., et al. 2004, ApJ, 607, 474
  57. [59]Ivans, I. I., Simmerer, J., Sneden, C., et al. 2006, ApJ, 645, 613
  58. [62]Johnson, J. A. & Bolte, M. 2001, ApJ, 554, 888
  59. [63]Just, O., Aloy, M. A., Obergaulinger, M., & Nagataki, S. 2022, ApJ, 934, L30
  60. [64]Kass, R. E. & Raftery, A. E. 1995, Journal of the American Statistical Association, 90, 773
  61. [65]Kobayashi, C., Karakas, A. I., & Lugaro, M. 2020, ApJ, 900, 179
  62. [66]Kobayashi, C., Mandel, I., Belczynski, K., et al. 2023, ApJ, 943, L12
  63. [67]Koutsouridou, I., Salvadori, S., Skúladóttir, Á., et al. 2026, arXiv e-prints, arXiv:2605.00990arxiv.org/abs/2605.00990
  64. [68]Koutsouridou, I., Skúladóttir, Á., & Salvadori, S. 2025, A&A, 699, A32
  65. [69]Lai, D. K., Bolte, M., Johnson, J. A., et al. 2008, ApJ, 681, 1524
  66. [70]Lattimer, J. M. & Schramm, D. N. 1974, ApJ, 192, L145
  67. [71]Lin, Y., Li, H., Jiang, R., et al. 2025, ApJ, 984, L43
  68. [72]Lind, K., Nordlander, T., Wehrhahn, A., et al. 2022, A&A, 665, A33
  69. [73]Lindgren, L., Bastian, U., Biermann, M., et al. 2021, A&A, 649, A4
  70. [74]Lucchesi, R., Jablonka, P., Skúladóttir, Á., et al. 2024, A&A, 686, A266
  71. [75]Magrini, L., Bensby, T., Randich, S., et al. 2026, arXiv preprint arXiv:2607.04916arxiv.org/abs/2607.04916
  72. [76]Marconi, A., Abreu, M., Adibekyan, V., et al. 2022, in Ground-based and Airborne Instrumentation for Astronomy IX, Vol. 12184, SPIE, 720–735
  73. [77]Mashonkina, L., Christlieb, N., Barklem, P. S., et al. 2010, A&A, 516, A46
  74. [78]Mashonkina, L., Christlieb, N., & Eriksson, K. 2014, A&A, 569, A43
  75. [79]Mashonkina, L. & Gehren, T. 2000, A&A, 364, 249
  76. [80]Mashonkina, L., Ryabtsev, A., & Frebel, A. 2012, A&A, 540, A98
  77. [81]Miller, J. M., Sprouse, T. M., Fryer, C. L., et al. 2020, ApJ, 902, 66
  78. [82]Molero, M., Magrini, L., Matteucci, F., et al. 2023, MNRAS, 523, 2974
  79. [83]Mösta, P., Roberts, L. F., Halevi, G., et al. 2018, ApJ, 864, 171
  80. [84]Nishimura, N., Sawai, H., Takiwaki, T., Yamada, S., & Thielemann, F.-K. 2017, ApJ, 836, L21
  81. [85]Nissen, P. E., Amarsi, A. M., Skúladóttir, Á., & Schuster, W. J. 2024, A&A, 682, A116
  82. [86]Nissen, P. E. & Schuster, W. J. 2010, A&A, 511, L10
  83. [87]Nissen, P. E. & Schuster, W. J. 2011, A&A, 530, A15
  84. [88]Patel, A., Metzger, B. D., Cehula, J., et al. 2025a, ApJ, 984, L29
  85. [89]Patel, A., Metzger, B. D., Goldberg, J. A., et al. 2025b, ApJ, 985, 234
  86. [90]Pian, E., D’Avanzo, P., Benetti, S., et al. 2017, Nature, 551, 67
  87. [91]Placco, V. M., Almeida-Fernandes, F., Holmbeck, E. M., et al. 2023, ApJ, 959, 60
  88. [92]Placco, V. M., Holmbeck, E. M., Frebel, A., et al. 2017, ApJ, 844, 18
  89. [93]Placco, V. M., Sneden, C., Roederer, I. U., et al. 2021, Research Notes of the American Astronomical Society, 5, 92
  90. [94]Reichert, M., Hansen, C. J., & Arcones, A. 2021a, ApJ, 912, 157
  91. [95]Reichert, M., Obergaülinger, M., Aloy, M. Á., et al. 2023, MNRAS, 518, 1557
  92. [96]Reichert, M., Obergaülinger, M., Eichler, M., Aloy, M. Á., & Arcones, A. 2021b, MNRAS, 501, 5733
  93. [97]Ricigliano, G., Jacobi, M., & Arcones, A. 2024, MNRAS, 533, 2096
  94. [98]Roederer, I. U., Beers, T. C., Hattori, K., et al. 2024, ApJ, 971, 158
  95. [99]Roederer, I. U., Hattori, K., & Valluri, M. 2018a, AJ, 156, 179
  96. [100]Roederer, I. U., Kratz, K.-L., Frebel, A., et al. 2009, ApJ, 698, 1963
  97. [101]Roederer, I. U., Lawler, J. E., Sobeck, J. S., et al. 2012, ApJS, 203, 27
  98. [102]Roederer, I. U., Preston, G. W., Thompson, I. B., et al. 2014, AJ, 147, 136
  99. [103]Roederer, I. U., Sakari, C. M., Placco, V. M., et al. 2018b, ApJ, 865, 129
  100. [104]Roederer, I. U., Sneden, C., Thompson, I. B., Preston, G. W., & Shectman, S. A. 2010, ApJ, 711, 573
  101. [105]Rosswog, S., Korobkin, O., Arcones, A., Thielemann, F.-K., & Piran, T. 2014, MNRAS, 439, 744
  102. [106]Sakari, C. M., Placco, V. M., Farrell, E. M., et al. 2018, ApJ, 868, 110
  103. [107]Saraf, P., Allende Prieto, C., Sivirani, T., et al. 2023, MNRAS, 524, 5607
  104. [108]Schatz, H., Becerril Reyes, A. D., Best, A., et al. 2022, Journal of Physics G Nuclear Physics, 49, 110502
  105. [109]Shah, S. P., Ezzeddine, R., Ji, A. P., et al. 2023, ApJ, 948, 122
  106. [110]Shejeelammal, J., Goswami, A., Goswami, P. P., Rathour, R. S., & Masseron, T. 2020, MNRAS, 492, 3708
  107. [111]Siegel, D. M., Barnes, J., & Metzger, B. D. 2019, Nature, 569, 241
  108. [112]Simonetti, P., Matteucci, F., Greggio, L., & Cescutti, G. 2019, MNRAS, 486, 2896
  109. [113]Siqueira Mello, C., Hill, V., Barbuy, B., et al. 2014, A&A, 565, A93
  110. [114]Skúladóttir, Á., Ernandes, H., Feuillet, D. K., et al. 2025, ApJ, 986, L21
  111. [115]Skúladóttir, Á., Hansen, C. J., Salvadori, S., & Choplin, A. 2019, A&A, 631, A171
  112. [116]Skúladóttir, Á., Puls, A. A., Amarsi, A. M., et al. 2023, Messenger, 190, 19
  113. [117]Skúladóttir, Á. & Salvadori, S. 2020, A&A, 634, L2
  114. [119]Sneden, C., Cowan, J. J., Lawler, J. E., et al. 2003, ApJ, 591, 936
  115. [121]Storm, N., Barklem, P. S., Yakovleva, S. A., et al. 2024, A&A, 683, A200
  116. [122]Storm, N., Bergemann, M., Fitner, P., et al. 2025, MNRAS, 538, 3284
  117. [123]Tolstoy, E., Hill, V., & Tosi, M. 2009, ARA&A, 47, 371
  118. [124]Vincenzo, F., Spitoni, E., Calura, F., et al. 2019, MNRAS, 487, L47
  119. [125]Wanajo, S., Fujibayashi, S., Hayashi, K., et al. 2024, Phys. Rev. Lett., 133, 241201
  120. [126]Wanajo, S., Sekiguchi, Y., Nishimura, N., et al. 2014, ApJ, 789, L39
  121. [127]Watson, D., Hansen, C. J., Selsing, J., et al. 2019, Nature, 574, 497
  122. [128]Xie, X.-J., Shi, J., Yan, H.-L., et al. 2024, ApJ, 970, L30
  123. [129]Xing, Q., Zhao, G., Aoki, W., et al. 2024, ApJ, 965, 79
  124. [130]Yong, D., Kobayashi, C., Da Costa, G. S., et al. 2021, Nature, 595, 223
  125. [131]Yuan, Z., Myeong, G. C., Beers, T. C., et al. 2020, ApJ, 891, 39

Paper details

vlnc:W_x4b3zk63
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents