From actinides to zinc: Using the full abundance pattern of the brightest star in Reticulum II to distinguish between different -process sites111This paper includes data gathered with the 6.5 m Magellan Telescopes located at Las Campanas Observatory, Chile.
The ultra-faint dwarf galaxy Reticulum II was enriched by a rare and prolific -process event, such as a neutron star merger. To investigate the nature of this event, we present high-resolution Magellan/MIKE spectroscopy of the brightest star in this galaxy. The high signal-to-noise allows us to determine the abundances of 41 elements, including the radioactive actinide element Th and first ever detections of third -process peak elements (Os and Ir) in a star outside the Milky Way. The observed neutron-capture element abundances closely match the solar -process component, except for the first -process peak which is significantly lower than solar but matches other -process enhanced stars. The ratio of first peak to heavier -process elements implies the -process site produces roughly equal masses of high and low electron fraction ejecta, within a factor of 2. We compare the detailed abundance pattern to predictions from nucleosynthesis calculations of neutron star mergers and magneto-rotationally driven jet supernovae, finding that nuclear physics uncertainties dominate over astrophysical uncertainties. We measure , somewhat lower than all previous Th/Eu observations. The youngest age we derive from this ratio is Gyr, indicating that current initial production ratios do not well describe the -process event in Reticulum II. The abundance of light elements up to Zn are consistent with extremely metal-poor Milky Way halo stars. They may eventually provide a way to distinguish between neutron star mergers and magneto-rotationally driven jet supernovae, but this would require more detailed knowledge of the chemical evolution of Reticulum II.
Alexander P. Ji
Ultra-faint dwarf galaxies (UFDs) are dwarf spheroidal galaxies with luminosities and metallicities (e.g., Kirby et al., 2008). They contain no gas (e.g., Westmeier et al., 2015) and have purely old stellar populations, forming most of their stars in the first Gyr of the universe (e.g., Brown et al., 2014; Weisz et al., 2014). Each UFD is the product of a short, independent burst of star formation and thus an ideal tool to investigate clean chemical enrichment events in the early universe.
About UFDs have been discovered within the virial radius of the Milky Way (Drlica-Wagner et al., 2015). The UFD Reticulum II (Ret II) was discovered in the Dark Energy Survey and quickly confirmed as a metal-poor UFD galaxy (, Bechtol et al., 2015; Simon et al., 2015; Koposov et al., 2015a, b; Walker et al., 2015). Surprisingly, the majority of the stars in Ret II displayed large enhancements of elements synthesized in the rapid neutron-capture process (-process), orders of magnitude higher than most other UFDs (, Ji et al., 2016a, c; Roederer et al., 2016, where Eu is a representative -process element), and similar to the most -process enhanced stars in the Milky Way stellar halo (or -II stars; Christlieb et al. 2004b; Beers & Christlieb 2005). It is thus clear that some sort of rare and prolific -process event enriched the system during its short, early period of star formation since all these stars also have low metallicities of .
The question remains about the origin of these -process elements. Ji et al. (2016a) estimated that such a rare and prolific event occurred only once out of every core-collapse supernovae, with each event producing of -process elements. This rate and yield clearly rule out -process production in neutrino-driven winds of ordinary core-collapse supernovae (Meyer et al., 1992; Woosley & Hoffman, 1992). Instead, they are consistent with expectations from a neutron star merger (NSM). In fact, after six decades of uncertainty regarding the astrophysical site of -process nucleosynthesis (Burbidge et al., 1957; Cameron, 1957), NSMs are now considered the favored site for -process nucleosynthesis. It has long been predicted that the ejecta released during a NSM have a very low electron fraction that easily synthesizes the heaviest -process elements (e.g., Lattimer & Schramm, 1974; Metzger et al., 2010; Goriely et al., 2011). The spectacular discovery of gravitational waves from the merging neutron star pair GW170817 and its electromagnetic counterpart SSS17a has confirmed that NSMs have red kilonova afterglows associated with the production of -process elements (Abbott et al., 2017a, b). Along with abundance measurements of plutonium in the ISM (Wallner et al., 2015; Hotokezaka et al., 2015), it now appears that NSMs dominate -process production in the universe today. A NSM origin for the -process elements in Ret II thus seems likely, and it would imply that NSMs can dominate -process production throughout cosmic history.
Indeed the UFD environment provides a way to circumvent the primary criticism of NSMs as the source of -process elements in metal-poor stars. While a rare -process site was needed to explain the large scatter in neutron-capture elements of halo stars (e.g., McWilliam et al., 1995), it was long thought NSMs could not fill this role as the delay time needed for a binary to coalesce through gravitational radiation would preclude NSMs from enriching metal-poor gas in the early universe quickly enough (e.g., Mathews & Cowan, 1990; Argast et al., 2004). However, the delay time is mitigated by inefficient/delayed star formation in a small galaxy like Ret II (Tsujimoto & Shigeyama, 2014; Ishimaru et al., 2015; Bland-Hawthorn et al., 2015; Ji et al., 2016a), as well as inhomogeneous metal mixing (Hirai et al., 2015; Shen et al., 2015). The main remaining challenge for the NSM interpretation in Ret II is velocity kicks that occur when forming the neutron stars, as these could remove the binary system from Ret II before it merges (Dominik et al., 2012; Bramante & Linden, 2016). Some models remedy this by proposing a population of neutron star binaries with low velocity kicks and rapid merging times (Beniamini et al., 2016).
Given current observations, the inferred rate and yield of the -process event in Ret II are also consistent with another proposed -process site, magneto-rotationally driven jet supernovae (MRDSNe). If some fraction of core-collapse supernovae have extremely high rotation speeds and magnetic fields, these special explosions could produce similar amounts of -process material to NSMs but without the delay time or velocity kicks (Wehmeyer et al., 2015; Cescutti et al., 2015). The primary concern in the literature appears to be whether such initial conditions can physically occur, since stellar evolution models have not been able to develop the high magnetic fields required (e.g., Mösta et al., 2017). Currently, all models of MRDSNe have the initial magnetic field as a free parameter of their initial conditions (e.g., Winteler et al., 2012; Nishimura et al., 2015), under the assumption that the magnetorotational instability (MRI) will amplify seed fields to the required strength. However, it is not yet clear if the MRI can actually reach the extremely high values required to actually synthesize -process elements (e.g., Rembiasz et al., 2016). Insufficient amplification prevents MRDSNe from synthesizing the heaviest -process elements (Nishimura et al., 2015, 2017). However only % of core-collapse supernovae would have to achieve these conditions to be nucleosynthetically relevant. This small fraction is not currently excluded by supernova observations (Winteler et al., 2012).
Investigation of more UFDs is likely to shed more light on this matter (e.g., Hansen et al., 2017), but another way to distinguish between NSMs and MRDSNe is precise detailed abundances of the heavy -process elements: the rare earth elements (such as La, Eu, Dy), third -process peak (such as Os, Ir, Au), and actinide elements (Th and U). Differences in the ejecta properties of NSMs and MRDSNe may lead to systematic differences in detailed abundance ratios of these heavy -process elements (Shibagaki et al., 2016; Kajino & Mathews, 2017). Indeed throughout the literature, nucleosynthesis calculations with NSMs and MRDSNe are unable to simultaneously reproduce the detailed isotopic abundance ratios of the extracted solar -process component, especially the rare earth elements and third -process peak (e.g., Winteler et al., 2012; Wanajo et al., 2014; Goriely et al., 2015; Just et al., 2015; Nishimura et al., 2015; Lippuner & Roberts, 2015; Wu et al., 2016). Most authors attribute these discrepancies to uncertainties in nuclear physics input (e.g., Kratz et al., 2014; Eichler et al., 2015; Mumpower et al., 2016; Nishimura et al., 2016). However, Shibagaki et al. (2016) proposed that this could be resolved if both NSMs and MRDSNe contributed heavy -process elements. Given the universality of the -process pattern, as seen in both in the sun and in metal-poor halo stars (e.g., Sneden et al., 2008), this seems like an unlikely solution. But in principle, it is possible that all -II halo stars observed so far formed from a composite population of -process sources. This can be resolved with further study of Ret II, which is thought to probe only one single -process event.
The relative abundance of actinide (Th and U) to stable -process elements is also still poorly understood. About of -process enhanced stars exhibit enhancements in Th, a so-called “actinide boost” (Mashonkina et al., 2014). Actinide elements are radioactive and contain isotopes with multi-Gyr half lives. A constraint on the age of the -process event can be placed by comparing the observed abundance to an initial production ratio. Thorium has been detected and age estimates have been made for many metal-poor halo stars (e.g., Sneden et al., 1996; Johnson & Bolte, 2001; Christlieb et al., 2004b; Frebel et al., 2007a; Ren et al., 2012; Mashonkina et al., 2014), but only once in a star outside the Milky Way (Aoki et al., 2007b).
Here, we present a high-resolution, high signal-to-noise optical spectrum of the brightest star in Ret II, with . We derive the abundance of 41 elements, including elements from the third -process peak and the actinide element thorium. In Section 2 we describe our observations and abundance analysis. We examine the -process pattern in Section 3, and compare it to nucleosynthesis predictions from NSMs and MRDSNe. In Section 4 we consider the Th abundance and the age of the -process event. In Section 5 we discuss the connection to the LIGO NSM event (GW170817), and how other elements like zinc may be a future path forward to distinguish between NSMs and other -process sites. We conclude in Section 6.
2 Observations and Abundance Analysis
We observed DES J033523540407 with the MIKE spectrograph (Bernstein et al., 2003) on the Magellan-Clay Telescope for a total of 22.1 hours on 2017 August , 2017 Aug , and 2017 October with the 07 slit. The weather was clear with seeing 07 for most of the observations. Data from each of the eight nights were reduced separately with the CarPy pipeline (Kelson, 2003). Subsequent data processing and abundance analysis was done with a custom analysis tool first described in Casey (2014). We normalized and coadded the eight spectra into one final normalized spectrum with for Å and for Å. Only wavelengths Å are useful. The approximate signal-to-noise per pixel is 70 at 4000 Å, 110 at 5200 Å, and 240 at 6500 Å, making this the highest signal-to-noise high-resolution spectrum of a UFD star ever taken. We measured heliocentric radial velocities from these and previous spectra using cross-correlation of the Mg b region (5150Å-5200Å) with a MIKE spectrum of CS22892052 as the template (see Table 1). The typical velocity uncertainty is km/s. Within this limit there is no evidence for binarity.
|2015 Oct 1-4||66.8 km/s||J16 (10 slit)|
|2017 Aug 13-16||67.1 km/s|
|2017 Aug 25||67.5 km/s|
|2017 Oct 9-11||66.8 km/s|
References. – S15 (Simon et al., 2015); SF11 (Schlafly & Finkbeiner, 2011); B15 (Bechtol et al., 2015); 2MASS (Skrutskie et al., 2006); A99 (Alonso et al., 1999); J16 (Ji et al., 2016c) Note. – S15 already included a reddening correction from SF11.
Note. – and are weighted mean and standard error. and are unweighted mean and standard deviation. is error from changes in stellar parameters. is the final uncertainty we adopt for each feature (see text). [X/Fe] ratios are calculated with Fe I or Fe II depending on the ionization state of X.
We performed a standard 1D LTE analysis using the -enhanced 1D plane-parallel model atmospheres from Castelli & Kurucz (2004) and the 2017 version of MOOG (Sneden, 1973), including the scattering routines from Sobeck et al. (2011)222https://github.com/alexji/moog17scat. Stellar parameters were determined through a combination of spectroscopic and photometric methods and summarized in Table 1. We first applied the procedure in Frebel et al. (2013)333We have verified the Frebel et al. (2013) temperature calibration remains valid for MOOG 2017 when scattering is included., resulting in , , , . For this bright star with many Fe lines, the statistical errors in stellar parameters are negligible so systematic errors dominate (150 K, 0.3 dex, 0.2 km s, 0.2 dex respectively; e.g., Ji et al. 2016a). This agrees within uncertainties of previous stellar parameter determinations ( K, , , ). We then used photometry from DES (Bechtol et al., 2015; Simon et al., 2015) and 2MASS with the appropriate reddening correction (Schlafly & Finkbeiner, 2011) and color-temperature relations (Alonso et al., 1999) assuming [Fe/H]. We obtain corresponding to . This matches our spectroscopic temperature, so we adopt K with a 50 K error dominated by intrinsic scatter in the temperature-color relation. We then derive photometrically (e.g. Mashonkina et al., 2017). DES J033523540407 has , with a bolometric correction of (Alonso et al., 1999) and distance modulus of (Bechtol et al., 2015; Koposov et al., 2015a). Assuming that and the star has , this results in . The corresponding microturbulence to balance abundance vs. line strength is . The spectroscopic has a much larger error bar, so we adopt the higher photometric . This of course causes a systematic LTE abundance difference between Fe I and Fe II of 0.17 dex. A NLTE correction increases the Fe I abundance by dex (Ezzeddine et al., 2017) and restores agreement between the Fe I and Fe II abundances. Such a correction is consistent with previous studies using NLTE corrections for Fe for stellar parameter determination (e.g., Mashonkina et al., 2017). Thus, whenever quoting [X/Fe] ratios for neutral or ionized species, we take care to consider ratios to the appropriate Fe abundance. We note that this does not affect our main results regarding the neutron-capture elements and their relative abundances, which only use the scale. Using either the Fe I or Fe II abundance as the model atmosphere metallicity also makes little difference in the final results. Our final model atmosphere parameters are K, , , , and .
We determined the abundance of O, Na, Mg, K, Ca, Ti, Cr, Fe, Ni, Zn, Y, Zr, Ce, Nd, Sm, Gd, Dy, and Er from equivalent width measurements of fitted Gaussian profiles. The abundance of C, N, Al, Si, V, Mn, Co, Sr, Ba, Mo, Ru, La, Pr, Eu, Tb, Ho, Tm, Yb, Hf, Os, Ir, and Th were measured with spectral synthesis. We used the solar -process isotope fractions for Ba and Eu (Sneden et al., 2008) and solar abundances from Asplund et al. (2009) whenever needed. Choices about which neutron-capture lines to measure were informed by examining the spectrum of HE 15230901 (Frebel et al., 2007a), and supplemented by data from Hill et al. (2002, 2017). Detailed synthesis line lists were then created based on software provided by Chris Sneden (priv. comm.). The software begins with the Kurucz (2011) line database and uses laboratory measurements from references in Sneden et al. (2009, 2014, 2016) to replace lines when possible. We additionally replaced the CH molecular lines with the list of Masseron et al. (2014). We included hyperfine structure and isotope splitting for Ba (McWilliam, 1998), Eu (Ivans et al., 2006), and Yb (Sneden et al., 2009). In principle, Nd, Sm, and Ir can show evidence for isotopic splitting (Cowan et al., 2005; Roederer et al., 2008). Our resolution and S/N are much too low to quantitatively detect shifts associated with isotopic differences, although for the Sm4424 line we find that the -process isotope ratios better fit the red wing of this feature compared to the -process isotope ratios. For many neutron-capture elements, only one or two lines can be measured. We show regions of the spectra around selected lines in Figure 1. The abundances and uncertainties of 41 elements and 5 upper limits in DES J033523540407 are presented in Table 2.
Given possible uncertainties due to atomic data, unknown blends, and NLTE or 3D effects, we adopt a minimum absolute uncertainty of 0.1 dex for all elements, and 0.2 dex for those elements measured with only a single line. However, for completeness, we also performed a comprehensive uncertainty analysis. Abundance precisions were derived for each individual line or feature representing the spectrum’s local data quality. For equivalent width measurements, we sampled 100 realizations of the best-fit Gaussian and continuum parameters and took a 68 percentile interval. For syntheses, we varied the element abundance until . The final element abundance () is an inverse variance weighted sum of individual features. The uncertainty is a quadrature sum of the statistical abundance precision (e.g., McWilliam et al., 2013) and the standard error of individual lines. For reference, we also provide in Table 2, the usual unweighted standard deviation reported by most high-resolution spectroscopic studies. Stellar parameter uncertainties were also propagated to abundance uncertainties, . Our formal total abundance uncertainty is the quadrature sum of and .
Our new abundance measurements are consistent with previous results of this star to within the expected errors (Ji et al., 2016c; Roederer et al., 2016). The high signal-to-noise of our spectrum allowed the determination of 16 new elements in this star: N, O, K, V, Zn, Mo, Ru, Tb, Ho, Er, Tm, Yb, Hf, Os, Ir, and Th. The typical abundance precision in dex has improved by a factor of from previous results. The abundance of light elements compared to halo stars of similar [Fe/H] is shown in Figure 2. The abundance of all light elements is perfectly in line with typical halo stars, so we do not discuss them further except for the CNO abundance in the next paragraph and the Zn abundance in 5.2 (see Nomoto et al. 2013; Frebel & Norris 2015 for detailed discussion of halo star abundance trends). The abundance pattern of neutron-capture elements compared to other -process stars is shown in Figure 3.
The determination of C, N, and O deserves some extra discussion. We detect two forbidden lines of O near 6300Å, resulting in a high abundance . These lines are normally too weak to see in metal-poor stars, but the large enhancement and our high S/N ( at 6300Å) allow this measurement. This abundance is high enough to influence molecular equilibrium, especially affecting carbon due to CO molecules. Using this O abundance, C is determined from CH regions near 4313Å and 4323Å. We measure C/C from several strong 13CH features between Å. The observed carbon abundance of is similar to the previous measurement in Ji et al. (2016c). Fixing this C abundance, is measured from the CN band near 3870Å.
In this paper we are concerned with the abundance of neutron-capture elements, so the observed C, N, and O abundances are mostly important insofar as they are blended with neutron-capture element lines. However, the origin of the enhanced CNO elements is of interest for e.g. understanding Population III stars. Unfortunately, a cool red giant like DES J033523540407 converts some C to N through internal mixing and CNO burning, though is relatively unaffected (e.g., Gratton et al., 2000). This same process is the reason the C/C ratio is relatively low in this star. We thus corrected the C abundance for evolutionary status of the star with models from Placco et al. (2014b). Using , the corrected carbon abundance is . If the spectroscopic is used instead, . Note that the initial N abundance should also be lower than observed and could be corrected assuming CN is constant. Unfortunately, the carbon correction model used here assumes that initially (V. Placco, priv. comm.). Given these uncertainties in the intrinsic C and N abundances, we just point out that metal-free Population III stars are typically expected to produce high amounts of CNO elements, either as faint supernovae (e.g, Umeda & Nomoto, 2002) or in spinstars (e.g., Frischknecht et al., 2016). These are unlikely to significantly affect the abundances of neutron-capture elements in DES J033523540407, as they produce negligible amounts of neutron-capture elements (see Ji et al. 2016b for a discussion).
Since the Th abundance is derived from a single strongly blended line and it is the only detected actinide, we discuss it in some detail here. The strong Th line at 4019Å is the only clearly detected Th feature in DES J033523540407. We show the best-fit abundance of A(Th) in Figure 1), which uses for Th (Nilsson et al., 2002). Unfortunately, this line is affected by several known blends: the blue side of the line is blended with Fe, Ni, Ce, and CH; the core of the line is blended with CH (this usually separates out at higher resolution); and the red wing is blended with Co. Thus, while a formal uncertainty from the spectrum is only dex, blends dominate the abundance uncertainty from this line. Unfortunately several of these blending features appear to have inaccurate atomic data. To well match the observed spectrum, we had to increase the strength of a Ce line at 4019.06Å by 0.3 dex to ; and the strength of a red Co feature at 4019.30Å by 0.8 dex to a total strength . Detailed examination of this region in other spectra (Frebel et al., 2007a; Hill et al., 2017) suggests that the high required Co is partly due to an unidentified feature(s) at or near 4019.25Å. These issues with (missing) atomic data have previously been noticed and required adjustments of similar magnitude (Morell et al., 1992; Sneden et al., 1996; Johnson & Bolte, 2001; Mashonkina et al., 2014). We verified that the changes to Ce and Co atomic data were also required to fit a high-resolution spectrum of HE 15230901.
At very high spectral resolution and signal-to-noise, these uncertainties only marginally affect the Th abundance ( dex, e.g. Sneden et al. 1996; Frebel et al. 2007a; Hill et al. 2017). For our data, we find that varying these two elements makes at most 0.1 dex difference. The Co change can thus be regarded as primarily cosmetic to achieve a good overall fit of the region containing the Th line. We compared our line list against all atomic data we could find in the literature (NIST; VALD, Kupka et al. 1999; Morell et al. 1992; Francois et al. 1993; Sneden et al. 1996; Johnson & Bolte 2001; Ren et al. 2012), finding that other atomic lines made minimal difference to the Th abundance. We also note another Ce line at 4019.47Å is clearly too strong in our linelist and in VALD, although it does not affect the Th abundance. Unfortunately, none of the Ce and Co lines we described here have recent laboratory measurements (Lawler et al., 2009, 2015). Varying the CH abundance by dex affects the Th abundance by . Accounting for all these uncertainties, we adopt the abundance A(Th). Using the same procedure, we verified that we can reproduce the Th abundance of HE 15230901 to within dex (Frebel et al., 2007a). We also detect a weak 4086Å Th feature in DES J033523540407. The abundance we derive is uncertain but consistent with the abundance of the line at 4019 Å. Other Th lines are undetectable or too blended.
Mo and Ru are detected in our spectrum but with quite uncertain abundances. Molybdenum has a feature at 3864.1Å that has been detected in the past but is highly blended with CN (Sneden et al., 2003; Ivans et al., 2006). We are able to measure a Mo abundance from this feature, but our line list does not fit the adjacent regions very well so we regard the derived abundance as quite uncertain. We also see evidence for nonzero abundances of two ruthenium lines near 3799Å and determine the abundance with a joint fit to both lines. These lines are in the wings of a Balmer line so also have a rather uncertain abundance. We adopt uncertainties of 0.3 dex for these elements.
We also searched for features of other elements: Cu, Ga, Rb, Rh, Sn, Pb, and U. There are no discernible features of any of these elements, so we calculate upper limits (corresponding to ). The upper limits are listed in Table 2. These upper limits only account for noise in the spectrum, not for uncertainties due to blends. Thus, the upper limits for Pb and U (calculated from the 4057Å and 3859Å features) should be taken with caution as they are significantly blended with CH and CN features. In the case of Sn at 3801Å, our line list does not fit the blending features well so we decided any upper limit would be unreliable.
3 A pure -process pattern
We plot the neutron-capture element abundance pattern of DES J033523540407 in Figure 3. For comparison, we show relative abundances of six well-studied -II stars (Sneden et al., 2008): HD221170 (Ivans et al., 2006), HD115444 (Westin et al., 2000), CS22892-052 (Sneden et al., 2003), HE 15230901 (Frebel et al., 2007a), BD 3248 (Cowan et al., 2002), CS31082-001 (Hill et al., 2002); as well as the solar -process component (Bisterzo et al., 2014). Each comparison is scaled to match the abundances of DES J033523540407 by minimizing the absolute residual of elements from (the rare earth elements). The bottom panel of Figure 3 shows the abundance difference between the stars and the solar -process pattern.
To clarify our subsequent discussion, we briefly remind the reader about some basics of the -process and the elements observable in optical spectra of metal-poor stars that probe different nucleosynthesis regimes. The observed -process pattern has three characteristic abundance peaks that result from three different closed neutron shells at (e.g., Burbidge et al., 1957; Sneden et al., 2008). According to nucleosynthesis calculations, the first peak (, ) is produced in -process ejecta with a relatively high electron fraction (e.g., Lippuner & Roberts, 2015). Metal-poor stars usually probe this element regime with Sr, Y, and Zr. All other -process elements are produced in ejecta with 444High ejecta can also make the heaviest elements if extremely high entropies increase the neutron-to-seed ratio, e.g., Woosley & Hoffman 1992; Farouqi et al. 2010; but these conditions are not achieved in current simulations of -process sites.. The second peak (, ) is best probed by Ba and La abundances, since elements directly in the peak (Te, I, Xe) are almost impossible to measure in stellar spectra. The third peak (, ) is most easily constrained with Os and Ir abundance measurements. There is a minor abundance peak corresponding to the rare earth elements, containing most of the stable lanthanides (, ). This is the region with the most robust abundance pattern and contains the prototypical -process element Eu that is measured in all -process metal-poor stars. The actinide region (, ) only manifests in the long-lived radioactive elements Th and U. Here, we have only mentioned the key elements most easily detected in optical spectra, but UV spectra principally allow detection of additional elements in or near the peaks (e.g., Roederer et al., 2012)
3.1 Comparison to the solar -process component and -II stars
We first consider the rare earth elements and the third -process peak (). Compared to previous measurements, we have now determined the abundance of five additional rare earth elements (Ho, Er, Tm, Yb, Hf) and two elements in the third -process peak (Os, Ir). This is the first time that any third peak elements have been measured in a star outside the Milky Way. The most striking aspect of our measurements is how closely they match the solar -process component and the abundances of other -II halo stars. As can be seen in Figure 3, the standard deviation of the residual of these 16 elements is only dex, similar to the typical abundance uncertainty.
The next clear feature in Figure 3 is that the abundance of the first -process peak elements (Sr, Y, Zr) in DES J033523540407 is lower than expected from the solar ratios by dex. Other -II stars also clearly display this deficiency, although there is significant scatter in the exact ratio with a standard deviation of dex (Sneden et al., 2008). Because of this, it is generally thought that the first peak elements can be produced independently from the heavier -process elements, in a different site, possibly in neutrino-driven winds of core-collapse supernovae (e.g., Travaglio et al., 2004; Montes et al., 2007; Honda et al., 2007; Arcones & Montes, 2011; Shibagaki et al., 2016).
As previously discussed in Ji et al. (2016c), Ret II has lower first peak abundances even compared to other -II stars. This is more clearly seen in the top three panels of Figure 4. Using the literature compilation of Abohalima & Frebel (2017), we identify 30 -II stars in the Milky Way halo and histogram their [Sr, Y, Zr/Eu] ratios (in black). Our star DES J033523540407 (red line with shaded red abundance uncertainty) clearly tends to lie towards the lower end of each distribution. This is consistent with the picture that Ret II probes a pure -process pattern from a single event, while some -II stars formed from gas that must have been significantly polluted by event(s) producing mostly neutron-capture elements in the first peak (presumably in core-collapse supernovae, e.g., Arcones & Montes 2011; Ji et al. 2016c). An alternate explanation is that the -process site intrinsically produces yields with some scatter in the relative amount of first peak and heavier -process elements (see Section 3.2.2).
3.2 Comparison to theoretical nucleosynthesis models
Here we examine if we can distinguish between different -process sites, namely the NSM and the MRDSN, based on the detailed -process abundance pattern. Since the pattern of DES J033523540407 so closely matches the solar abundance pattern, much of our subsequent comparisons and discussion has already been considered individually by the nucleosynthesis modelers (e.g., Wanajo et al., 2014; Eichler et al., 2015; Goriely et al., 2015; Nishimura et al., 2015, 2017; Wu et al., 2016; Radice et al., 2016; Shibagaki et al., 2016). However, here we aim to bring together the most salient features from an observational perspective, with the added insight that the -process pattern in Ret II probes a single event.
In Figure 5, we compare the abundance pattern of DES J033523540407 to nucleosynthesis calculations of the -process in the dynamically cold ejecta during a neutron star merger (Eichler et al., 2015), in a neutron star merger disk wind (Wu et al., 2016), and in a magneto-rotationally driven jet supernovae (Nishimura et al., 2017). Overall, there is remarkably good agreement between the models, DES J033523540407, and the solar -process component. This underscores the robustness of the basic -process nuclear physics (e.g., decay from closed neutron shells, fission cycling), as well as the success of much research aiming to reproduce the detailed isotopic ratios of the solar -process component. Note that the predictions for the radioactive actinides Th and U are abundances after initial production, and have not been adjusted for multiple Gyrs of radioactive decay.
3.2.1 Description of -process site models
Neutron star mergers have two main classes of ejecta: dynamical/prompt ejecta, and wind/post-merger ejecta. The cold dynamical NSM ejecta model (Eichler et al., 2015) tracks the traditional tidal ejecta that synthesize the heaviest -process elements due to its extremely low electron fraction (, e.g., Lattimer & Schramm 1974; Lattimer et al. 1977). However, as is apparent in Figure 5, these ejecta are so neutron-rich that it produces only negligible amounts of elements from the first -process peak. More recent calculations have shown that including shock heated ejecta from the NS collision and weak neutrino interactions can greatly increase the in some parts of the dynamical ejecta (e.g., Wanajo et al., 2014; Radice et al., 2016). Whether this is sufficient to reproduce the observed first peak elements appears to depend on the treatment of neutrino transport in the simulations.
An alternate means of ejecting matter in NSMs is in disk winds, following the prompt dynamical ejecta (e.g., Fernández & Metzger, 2013; Just et al., 2015). Here, neutron-rich material re-coalesces into a disk around the merger remnant. Disk winds develop through a combination of viscous heating and nuclear heating from -particle formation. These winds can actually eject more mass than the dynamical ejecta (Wu et al., 2016). Weak force interactions also greatly increase the of the disk material, resulting in a full distribution of -process elements, as seen in Figure 5. For illustration, we also add together the NSM disk wind and dynamical ejecta in Figure 5 to emphasize that the nucleosynthetic signature of a NSM probably contains a superposition of both types of ejecta.
Magneto-rotationally driven supernovae have a different explosion mechanism than standard core-collapse supernovae. Rather than being driven by neutrino heating, in these models high magnetic pressure launches jets of material out along the rotational axis (Takiwaki et al., 2009). In current calculations of MRDSNe, the initial magnetic fields and rotation velocities are not computed self-consistently from stellar evolution, but are instead set to values that will induce explosions, e.g. the iron core rotates at of breakup speed with a magnetic field of G (Winteler et al., 2012; Nishimura et al., 2015). Recent work has argued that even higher magnetic fields of G are required for this mechanism to work (Mösta et al., 2017). When the rotation speed and magnetic field are sufficiently high, a low jet is launched promptly and can undergo full -process nucleosynthesis (prompt jet model in Figure 5). Otherwise, a jet takes some time to form, causing the of the ejecta to be higher so the nucleosynthesis only proceeds to the first -process peak (delayed jet model in Figure 5). The abundance pattern of the delayed jet model is qualitatively similar to current expectations for a neutrino-driven wind in a core-collapse supernova (e.g., Arcones & Montes, 2011; Wanajo, 2013). Since right now the initial conditions are put in by hand, the of MRDSN ejecta are essentially a free parameter that depends on the relative amount of magnetic energy vs. neutrino heating (this is made explicit in Nishimura et al. 2017). However, this principally also allows MRDSN to produce the full range of -process nucleosynthesis patterns (Nishimura et al., 2017).
3.2.2 The first -process peak
Comparing the four models in Figure 5, the most obvious effect on the abundances is how the distribution of ejecta drastically affects the ratio of first peak elements to the heavier -process elements. In fact, there is a rather sharp cutoff where almost all ejecta with synthesize just the first peak elements, while almost all low ejecta synthesize the heavier -process elements (Lippuner & Roberts, 2015). Thus, since the cold dynamical NSM ejecta all have , it produces almost none of the elements with (Korobkin et al., 2012). On the opposite end, the MRDSN delayed jet ejecta have , so almost none of the heavier -process elements are formed. The NSM disk wind (Wu et al., 2016) and the MRDSN prompt jet (Nishimura et al., 2015, 2017) both produce a distribution of that ranges from , allowing them to synthesize elements from both regions.
It is thus clear that the ratio of ejecta with to ejecta with in the -process event () directly manifests as the ratio of first peak elements to heavier -process elements, (). To determine this value in DES J033523540407, we assume elements from to are associated with the first peak (), while heavier elements with belong to the second and third peak (). The ratio of first peak elements to the main -process pattern in Ret II is the ratio expected from the Bisterzo et al. (2014) solar -process pattern (Figure 3). Observed abundances are number densities, so we convert them into masses using average atomic masses from the solar -process isotope distribution (Bisterzo et al., 2014). We also use the solar -process isotopes to fill in the mass of unmeasured elements. Mathematically, this corresponds to
where is the mean mass of the element with proton number and is the number density of element from Bisterzo et al. (2014) (i.e., the black line in Figure 3). Note that includes the long-lived actinides Th and U, but these contribute only 0.3% to so are unimportant. The resulting value is for DES J033523540407555Calculating the same ratio for the Bisterzo et al. (2014) -process component gives , while the classical -process (Arlandini et al., 1999; Simmerer et al., 2004) gives . Note that changing the solar pattern makes no difference to for DES J033523540407 because the factor of also has to be adjusted accordingly..
Astrophysical -process sites have many parameters for which the overall of their ejecta can be adjusted (e.g., binary mass ratios, disk masses, or neutrino irradiation in NSMs; strength of magnetic field or rotation in MRDSNe). It thus seems very likely that the -process site should have some intrinsic scatter in . We can estimate an upper limit on the amount of intrinsic scatter by looking at the whole population of -II halo stars. Figure 4 shows that the observed [Sr, Y, Zr/Eu] ratios have a range of dex. Applying this range to Equation 1 corresponds to . Thus, the -process site must produce a fairly robust mass ratio of ejecta with high and low , i.e. equal to within a factor of .
We highlight that this intrinsic scatter is an upper limit, because the stars with larger may be contaminated by a separate site producing only first peak elements from another site (see Section 3.1). In fact, we know such contamination also exists in Reticulum II, due to the nonzero Sr abundance measured in one of the most metal-poor Ret II stars that does not have -process enhancement: Roederer et al. (2016) find for the star DES J033531540148. This amount of contaminating material is of that found in the -process enhanced stars (Ji et al., 2016c), so it is well below our measurement precision and does not significantly impact our inferred ratio of .
The small intrinsic scatter in inferred from -process stars is in stark contrast to the mass ratios of high and low ejecta in many simulations, where differences in neutrino treatment cause orders of magnitude differences in the amount of high and low ejecta (e.g., Wanajo et al., 2014; Nishimura et al., 2015; Radice et al., 2016; Roberts et al., 2017). Of course our calculation is only a rough estimate that cannot replace a full nucleosynthesis network calculation, but it underscores the fact that the relative abundance scatter among -II halo stars place a fairly stringent constraint on the distribution of -process events. We conclude there must be some underlying physical explanation for why the distribution of -process events is so robust.
3.2.3 The heaviest -process elements
There are some more subtle discrepancies with the rare earth elements. Two especially notable abundance differences of size dex that are well beyond our uncertainties: (1) both the dynamical and disk wind NSM ejecta produce too low abundances of Os and Ir, and (2) the MRDSN underproduces elements from La through Nd. These are indicated in shaded gray regions in Figure 5. They are similar to the deficiencies in these sites, as discussed in Shibagaki et al. (2016). However, it is clear that both discrepancies have to be resolved by the same -process site, since Ret II is unambiguously enriched by a single -process event. This rules out the multi-site solution proposed by Shibagaki et al. (2016) to explain the solar -process isotope ratios.
It is tempting to also use these discrepancies to constrain properties of the astrophysical site (e.g., the Os and Ir discrepancy can be reduced in the NSM disk wind model by varying the disk mass; Wu et al. 2016). However, a more likely explanation is that these discrepancies are results of nuclear physics uncertainties in -process nucleosynthesis networks. Indeed, since the DES J033523540407 pattern matches the solar -process pattern so well, nuclear physics solutions for these two types of abundance discrepancies have already been offered. Problem (1) appears to arise because of extra neutrons late in the -process (Eichler et al., 2015). This can be resolved with updated -decay rates of isotopes near the third -process peak (Eichler et al., 2015; Nishimura et al., 2016; Marketin et al., 2016) or different nuclear mass models (Mendoza-Temis et al., 2015). Problem (2) is related to the stability of isotopes slightly offset from the closed neutron shells (Shibagaki et al., 2016). This is alleviated by including more refined fission fragment distributions or newer nuclear mass models (Kratz et al., 2014; Eichler et al., 2015; Shibagaki et al., 2016; Nishimura et al., 2017).
It thus seems that until the nuclear physics input is better constrained, astrophysical sites cannot be distinguished by examining the detailed distribution of elements from the second to third -process peak. Overall though, Figure 5 shows that observational precision of neutron-capture element abundances in metal-poor stars are already able to distinguish between some predictions from -process sites. Improved modeling or measurements of fundamental nuclear parameters are needed before such comparisons can be used to understand astrophysical sites (e.g., Mumpower et al., 2016; Nishimura et al., 2016). Fortunately large experimental efforts, e.g. the Facility for Rare Isotope Beams (FRIB) are underway to tackle at least some of these critical issues. In the coming decades, this should lead to significant reductions of uncertainties although the heaviest neutron-rich isotopes will remain unreachable (e.g., Arcones et al., 2016; Mumpower et al., 2016; Kajino & Mathews, 2017). In the meantime, examining the abundance differences in the first peak and actinides may be more useful for understanding the astrophysical site of the -process.
4 The actinide element thorium
Since Th is radioactive with a half life of 14.05 Gyr, our Th abundance provides a way to date the production of the -process elements in Ret II by comparing to the abundance of other stable -process elements and an initial production ratio. Here we use Eu as a representative stable -process element. Using other rare earth elements makes little difference, since they all closely match the -process pattern. We measure for DES J033523540407. Our 0.24 dex uncertainty estimate is quite conservative, including spectrum noise (0.04 dex), stellar parameter uncertainties (0.10 dex), blends with other elements (0.2 dex), and Eu abundance uncertainty (0.05 dex). To our knowledge, this is only the second time Th has been detected in a galaxy other than the Milky Way (the other being Ursa Minor, Aoki et al. 2007b).
4.1 Comparison to other metal-poor stars
Figure 6 shows for DES J033523540407 (in red), eight -II stars (Abohalima & Frebel, 2017, solid black histogram), and from a large sample of 41 Th/Eu measurements in general metal-poor stars (Ren et al., 2012, dashed blue histogram). We note that the sample from Ren et al. (2012) is quite inhomogeneous, including stars from and . It includes two stars with , which are both marginal Th measurements (detected at only ). We also plot the Th/Eu ratio and uncertainty for the UMi star COS 82 (Aoki et al., 2007b). We note this abundance is derived from the Th5989Å line rather than the Th4019Å line that is used for every other star. Even with our large uncertainty, it is immediately clear that the Th/Eu ratio in DES J033523540407 is very extreme and thus falls at the lower end of all observed distributions.
This is not the first time an extreme Th/Eu ratio has been found. CS31082-001 (Hill et al., 2002) was the first “actinide boost” star, in that it had unusually high Th. Since then, several actinide boost stars have been discovered, with . It is now thought that a large fraction of -process stars may be actinide boosted (e.g., Mashonkina et al. 2014 found six actinide boost stars out of 18 -process enhanced halo stars). The origin of the actinide boost remains unknown (Hill et al., 2017).
Given the overall small number of Th measurements, it might be possible that there is also a separate population of “actinide deficient” stars, of which DES J033523540407 is the first example. Since Th is hard to detect, it would not be surprising if lower-Th stars were mostly not identified (e.g., the two stars in Ren et al. 2012 with low Th/Eu are only detections and often would not be reported). Of course, given the large uncertainty, DES J033523540407 is also only away from having a normal Th/Eu ratio.
4.2 Dating the -process event
The age of the -process event can be derived by
where is some stable -process element such as Eu, is an initial production ratio (PR) from a theoretical -process calculation, and is the observed abundance. This equation is easily derived from the 14.05 Gyr half life of Th. The long half life implies that any date measured from Th is very sensitive to small abundance changes, i.e. a 0.01 dex abundance difference results in 0.47 Gyr age difference and a 13 Gyr age difference causes only 0.28 dex decrease in Th abundance.
However, the current key challenge for dating an -process event is what PR should be used for . PRs used in the literature have been derived from a site-independent “waiting point” method (Schatz et al., 2002; Kratz et al., 2007) or from a high entropy neutrino wind model (Farouqi et al., 2010). Predictions range from to (see summaries in Placco et al. 2017; Hill et al. 2017), which already imply Gyr of systematic uncertainty. Using these PRs, the -process event in Ret II occurred to Gyr ago, with a combined uncertainty of 11.2 Gyr (2.8 Gyr statistical uncertainty, 10.3 Gyr systematic uncertainty that is dominated by the CH blends). This is clearly higher than the 13.8 Gyr for the age of the universe, as expected from cosmology (Planck Collaboration et al., 2016). The discrepancy could possibly be attributed to our large observational uncertainty, including deriving Th/Eu from a single Th line. However, another likely systematic issue at hand is that the available PRs simply do not apply to the case of Reticulum II. The waiting point method and the neutrino wind may not well describe the NSMs or MRDSNe scenarios. Instead, the models shown in Figure 5, at face value, predict PRs that range from to . Unfortunately, these higher PRs actually increase the tension with the age of the universe. However, there are still significant uncertainties in the nuclear physics (e.g., different mass models) that can affect the calculated Th/Eu ratio by up to a factor of 10 (M. Eichler & M.-R. Wu, priv. comm.). Hence, any of these values should be taken with caution without further investigation.
To guide future theoretical studies, we instead invert the age determination to find a range of Th/Eu PRs consistent with Ret II. Assuming the -process event happened in Ret II Gyr ago (the age of Ret II from its color-magnitude diagram, Bechtol et al. 2015), the Th/Eu PR would have to be .
Ultimately, any interpretation of the Th abundance in Ret II is limited by its relatively large abundance uncertainty. Improving the Th abundance in this star can probably only be achieved by obtaining an even higher resolution spectrum of similar or better signal-to-noise, but the signal-to-noise and resolution achieved on this star is near the limit of what can be reasonably done for a UFD star using current facilities. However, such a measurement will be easily accomplished with a high-dispersion spectrograph on a 30 m class telescope (e.g. G-CLEF, Szentgyorgyi et al. 2016). A larger telescope would also allow detailed study of fainter stars in Ret II that have lower carbon abundances. This might allow determination of other key elements like U and Pb. U is a much better probe for the age of the -process event, since U has a shorter half life of 4.5 Gyr, and Th/U ratios are probably more robust to uncertainties in nuclear physics. Unfortunately U is extremely difficult to measure, and only four stars in the literature have U detections (Hill et al., 2002, 2017; Frebel et al., 2007b; Placco et al., 2017).
A related question is why most -process halo stars have fairly different Th/r ratios compared to DES J033523540407. One possibility is that all previously discovered halo -process stars have actually been enriched by multiple -process events, thus raising their Th abundance relative to their stable element abundances and that found in Ret II. This seems unlikely to us, given that a substantial fraction of metal-poor halo stars, including -process stars, might stem from dwarf galaxies (e.g., Frebel et al., 2010). However, the overall number of metal-poor stars with Th abundances is still quite low and more observations are needed to draw firm conclusions. The other option is that the -process is not universal for the actinide elements, i.e. that there is intrinsic scatter in the production ratio of actinides to rare earth elements. In fact the existence of actinide boost stars already implies that no single universal production ratio can explain the Th/Eu ratio in all -process halo stars (Hill et al., 2002). The fairly broad Th/Eu distribution from Ren et al. (2012) (bottom panel of Figure 4) also suggests that there could be a continuum of Th/Eu ratios produced in the -process. If so, dating -process stars with Th requires fitting nucleosynthesis models to the abundances of each individual star in order to predict production ratios (this is the approach taken by Hill et al., 2017). Additional theoretical work is needed to understand the origin of intrinsic actinide scatter, and this can be aided by more measurements of thorium abundances in metal-poor stars.
5.1 Comparison to GW170817/SSS17a
The recent discovery of the binary neutron star merger GW170817/SSS17a has provided the first direct constraints on -process yields from NSMs. Somewhat surprisingly, there were two components detected in the afterglow (e.g., Drout et al., 2017). About of ejecta followed the standard expectation of a faint, long-lasting, red afterglow flung out at , and requiring significant amounts of lanthanides (e.g., Barnes & Kasen, 2013). This mass was significantly larger than most expectations from models (e.g., Wu et al., 2016). It required either a significant amount of mass from a fast disk wind (Margalit & Metzger, 2017) or an asymmetric mass ratio (Kilpatrick et al., 2017). There was also early fast-moving () blue emission that is mostly interpreted as of lanthanide-free ejecta (e.g., Drout et al., 2017; Metzger, 2017), though an alternate explanation is a shock breakout from a cocoon (Piro & Kollmeier, 2017; Kasliwal et al., 2017).
Assuming that -process rich metal-poor stars such as those found in Ret II are probing the ejecta of events like GW170817, the detailed abundances of these stars can provide some additional insights about the nature of the event (also see Côté et al. 2017). Most importantly, using the ratio of the first peak elements to the rest of the -process elements from -II halo stars (Section 3.2.2), we know that the mass of ejecta with and must be equal to within a factor of (with a point estimate of from Ret II). If we naively assume that the blue emission from GW170817 is all ejecta and the red emission is all ejecta, the mass ratio of appears to contradict our expected ratio of . However, the red emission only requires mass fraction of lanthanides (Drout et al., 2017), while the solar -process pattern has a lanthanide mass fraction of when including the first peak (Bisterzo et al., 2014). Thus, there is clearly some higher ejecta mixed into the red component of GW170187, as implied by the fact that all -II stars have .
Finally, we compare the yield and rate of an -process event expected for Ret II to those values inferred from GW170817. Ji et al. (2016a) estimated the -process event in Ret II produced . Typical total ejecta masses for GW170817 are (Drout et al., 2017; Metzger, 2017). Côté et al. (2017) estimated that this would turn into of Eu (Côté et al., 2017). The yields of the two events thus appear to be in agreement. Ji et al. (2016a) also estimated that one NSM occurred every core-collapse supernovae. The currently inferred binary neutron star merger rate is Gpc yr (Abbott et al., 2017a), while a core-collapse supernova rate is Gpc yr (Taylor et al., 2014). Thus, the point estimate of is , much higher than what was expected from population synthesis models (Belczynski et al., 2017), and is expected from Ret II and other UFDs. It is possible the rate estimate in Ji et al. (2016a) should be amended to include the mildly -process enhanced galaxy Tuc III (Hansen et al., 2017). This galaxy could have received its chemical signature from an off-center NSM explosion (Safarzadeh & Scannapieco, 2017). But this would still imply a rate of 1 event every SNe. However, the rate of NSMs detected in UFDs would be lower than the actual NSM rate if a significant fraction of the NSMs are ejected from UFDs (Bramante & Linden, 2016).
5.2 Zinc as a constraint on the origin of the -process elements
Since uncertain nuclear physics appears to be the dominant source of uncertainty in the predictions of -process abundances, we cannot currently use the observed -process abundance pattern to distinguish between NSMs and MRDSNe as the source of -process elements in Ret II (see Figure 5). However, while NSMs produce almost exclusively -process elements, MRDSNe will also synthesize many other lighter elements as part of the explosion. It is thus possible that these elements will leave an additional imprint that might differentiate the -process site.
One of the most promising elements for this purpose is zinc. Nishimura et al. (2017) recently noted that MRDSNe appear to generically produce a very high ratio. Their fiducial -process model has . We measure in DES J033523540407 based on two clean lines at 4722Å and 4810Å. Roederer et al. (2016) also measured in another Ret II star (DES J033607540235). At face value, these are much lower [Zn/Fe] ratios than expected for a MRDSN, but we must consider the fact that the MRDSN ejecta are being added to an ISM that has already been enriched in metals by regular core-collapse supernovae (CCSNe). Indeed, the MRDSNe eject only small amounts of mass ( total metals, i.e., Fe, Zn, Eu; Nishimura et al. 2017) compared to typical CCSNe.
We quantify how the Zn abundance of Ret II would be impacted by adding a single MRDSN yield to its gas in Figure 7. Ideally, we would add the yields to a direct measurement of the Zn abundance in Ret II as it would have been prior to any -process event. This is in principle possible as there are two metal-poor stars known in Ret II without any -process enhancements, but those stars do not have a Zn measurement yet (Ji et al., 2016c; Roederer et al., 2016). Instead, we fit [Zn/Fe] for halo stars (black points and black line in Figure 7; Abohalima & Frebel 2017) and use this as a reference point. The blue lines in Figure 7 show the expected [Zn/Fe] enhancement after adding the yield of one MRDSN. The expected trend depends critically on how much H gas the MRDSN yield dilutes into: less dilution corresponds to a stronger enhancement. If the MRDSN ejecta are diluted into of gas in Ret II, there should be a noticeable enhancement in [Zn/Fe]. This is not observed in DES J033523540407 and the other star with a Zn measurement. However, if instead the yield is more diluted, then it would not be possible to distinguish a MRDSN from a NSM using [Zn/Fe]. A erg SN explosion dilutes into only of gas, but our expectation is that turbulent mixing in a UFD between the supernova and subsequent star formation would usually increase this to of gas (Ji et al., 2015, 2016a). Assuming it is a MRDSN, we can use the predicted Eu yield and the observed [Eu/H] to infer that the mixing mass is . Without additional investigations of metal mixing in UFDs, we thus cannot clearly distinguish between NSMs and MRDSNe in Ret II using the Zn abundance.
If MRDSNe are the overall dominant source of -process elements in the early universe, one might also expect a correlation between Eu-enhancement and Zn in metal-poor stars. Using a large homogeneous sample of halo star abundances, Roederer et al. (2014a) found no observed correlation between extreme -process enhancement and the abundance of any element with . For Zn, the maximum allowed correlation was 0.1 dex. This might then be considered tentative evidence that MRDSNe are not responsible for most -process enhanced stars. But again, the conclusion depends on the expected dilution mass, and thus the gas from which a random -process-enhanced halo star formed.
We furthermore note that, in general, Zn abundances are difficult to interpret given current knowledge of Zn nucleosynthesis and its production during complete Si-burning. Metal-poor halo stars exhibit increasing [Zn/Fe] with decreasing [Fe/H]. Standard spherical metal-poor CCSNe models underproduce the observed [Zn/Fe] trend (Nomoto et al., 2013), while high energy hypernovae do produce but underproduce -elements (Umeda & Nomoto, 2002). Collimated jet-like outflows in supernovae (analogous to the delayed jet model of Nishimura et al. 2015) have been proposed as a way to simultaneously produce high Zn and high -elements (Tominaga, 2009), but these would have to be a generic feature of low [Fe/H] supernovae to explain the overall observed trend. Complicating this is the fact that electron-capture supernovae may also produce very high [Zn/Fe] ratios (Hirai et al., 2018; Wanajo et al., 2018). Regardless, a detailed chemical evolution model of Ret II will open up additional paths to understanding the nature of the -process and early nucleosynthesis, and fulfill the promise of dwarf galaxy archaeology (Frebel & Bromm, 2012).
We present a detailed abundance study of the brightest star in the -process UFD galaxy Reticulum II, DES J033523540407. This is the highest signal-to-noise spectrum of a UFD star taken to date, allowing us to measure the abundance of 18 elements up to Zn and 23 neutron-capture elements.
We add 11 neutron-capture elements to the -process pattern of DES J033523540407, which establishes the universal -process pattern from Ba through Ir for the first time outside the Milky Way (, ; Figure 3, Section 3.1). As with other -process stars in the Milky Way, the abundances of first peak elements (Sr, Y, Zr, Mo, Ru) are systematically low compared to both the solar -process component as well as other -II stars (Figure 4). Using this, we infer that the -process site must produce roughly equal masses of ejecta with and ejecta with (Section 3.2.2). This constraint on the amount of neutron-rich and neutron-poor ejecta is broadly consistent with the neutron star merger event associated with GW170817 (Section 5.1).
We also compare our observed pattern to detailed nucleosynthesis models of neutron star mergers and magneto-rotationally driven jet supernovae (Figure 5). We show that a single -process site produces both the rare earth elements and the third -process peak, disproving previous suggestions that multiple -process sites are needed to reproduce these features in the solar -process pattern. Improvements in nuclear physics inputs are needed before observations of second through third -process peak elements can be used to further constrain astrophysical sites (Section 3.2.3). Until then, abundances of the first peak elements and actinides display more variance and are currently a more useful constraint on the astrophysical sites. We note that detailed chemical evolution modeling of Ret II may allow the use of other elements like Zn as a constraint on the -process site (Section 5.2).
The Th abundance in this star is , implying . This is one of the lowest Th/Eu ratios observed so far (Figure 6, Section 4.1). Age estimates based on comparing the observed ratio to theoretical Th/Eu initial production ratios from the literature suggest ages older than 15 Gyr, though with a large total uncertainty of Gyr (Section 4.2). It appears more theoretical work regarding production ratios in different -process sites is needed to understand the Th abundance of DES J033523540407and the -process event that occurred in Ret II.
- Abbott et al. (2017a) Abbott, B. P., Abbott, R., Abbott, T. D., et al. 2017a, Physical Review Letters, 119, 161101
- Abbott et al. (2017b) —. 2017b, ApJ, 848, L12
- Abohalima & Frebel (2017) Abohalima, A., & Frebel, A. 2017, ArXiv e-prints, arXiv:1711.04410
- Allen et al. (2012) Allen, D. M., Ryan, S. G., Rossi, S., Beers, T. C., & Tsangarides, S. A. 2012, A&A, 548, A34
- Alonso et al. (1999) Alonso, A., Arribas, S., & Martínez-Roger, C. 1999, A&AS, 140, 261
- Aoki et al. (2007a) Aoki, W., Beers, T. C., Christlieb, N., et al. 2007a, ApJ, 655, 492
- Aoki et al. (2010) Aoki, W., Beers, T. C., Honda, S., & Carollo, D. 2010, ApJ, 723, L201
- Aoki et al. (2007b) Aoki, W., Honda, S., Sadakane, K., & Arimoto, N. 2007b, PASJ, 59, L15
- Aoki et al. (2002) Aoki, W., Ryan, S. G., Norris, J. E., et al. 2002, ApJ, 580, 1149
- Aoki et al. (2005) Aoki, W., Honda, S., Beers, T. C., et al. 2005, ApJ, 632, 611
- Aoki et al. (2008) Aoki, W., Beers, T. C., Sivarani, T., et al. 2008, ApJ, 678, 1351
- Arcones & Montes (2011) Arcones, A., & Montes, F. 2011, ApJ, 731, 5
- Arcones et al. (2016) Arcones, A., Bardayan, D. W., Beers, T. C., et al. 2016, ArXiv e-prints, arXiv:1603.02213
- Argast et al. (2004) Argast, D., Samland, M., Thielemann, F.-K., & Qian, Y.-Z. 2004, A&A, 416, 997
- Arlandini et al. (1999) Arlandini, C., Käppeler, F., Wisshak, K., et al. 1999, ApJ, 525, 886
- Asplund et al. (2009) Asplund, M., Grevesse, N., Sauval, A. J., & Scott, P. 2009, ARA&A, 47, 481
- Astropy Collaboration et al. (2013) Astropy Collaboration, Robitaille, T. P., Tollerud, E. J., et al. 2013, A&A, 558, A33
- Barbuy et al. (2005) Barbuy, B., Spite, M., Spite, F., et al. 2005, A&A, 429, 1031
- Barklem et al. (2005) Barklem, P. S., Christlieb, N., Beers, T. C., et al. 2005, A&A, 439, 129
- Barnes & Kasen (2013) Barnes, J., & Kasen, D. 2013, ApJ, 775, 18
- Bechtol et al. (2015) Bechtol, K., Drlica-Wagner, A., Balbinot, E., et al. 2015, ApJ, 807, 50
- Beers & Christlieb (2005) Beers, T. C., & Christlieb, N. 2005, ARA&A, 43, 531
- Behara et al. (2010) Behara, N. T., Bonifacio, P., Ludwig, H.-G., et al. 2010, A&A, 513, A72
- Belczynski et al. (2017) Belczynski, K., Askar, A., Arca-Sedda, M., et al. 2017, ArXiv e-prints, arXiv:1712.00632
- Beniamini et al. (2016) Beniamini, P., Hotokezaka, K., & Piran, T. 2016, ApJ, 829, L13
- Bernstein et al. (2003) Bernstein, R., Shectman, S. A., Gunnels, S. M., Mochnacki, S., & Athey, A. E. 2003, Proc. SPIE, 4841, 1694
- Bisterzo et al. (2014) Bisterzo, S., Travaglio, C., Gallino, R., Wiescher, M., & Käppeler, F. 2014, ApJ, 787, 10
- Bland-Hawthorn et al. (2015) Bland-Hawthorn, J., Sutherland, R., & Webster, D. 2015, ApJ, 807, 154
- Bonifacio et al. (2012) Bonifacio, P., Sbordone, L., Caffau, E., et al. 2012, A&A, 542, A87
- Bonifacio et al. (2009) Bonifacio, P., Spite, M., Cayrel, R., et al. 2009, A&A, 501, 519
- Bramante & Linden (2016) Bramante, J., & Linden, T. 2016, ArXiv e-prints, arXiv:1601.06784
- Brown et al. (2014) Brown, T. M., Tumlinson, J., Geha, M., et al. 2014, ApJ, 796, 91
- Burbidge et al. (1957) Burbidge, E. M., Burbidge, G. R., Fowler, W. A., & Hoyle, F. 1957, Reviews of Modern Physics, 29, 547
- Burris et al. (2000) Burris, D. L., Pilachowski, C. A., Armandroff, T. E., et al. 2000, ApJ, 544, 302
- Çalışkan et al. (2014) Çalışkan, Ş., Caffau, E., Bonifacio, P., et al. 2014, A&A, 571, A62
- Caffau et al. (2011) Caffau, E., Bonifacio, P., François, P., et al. 2011, The Messenger, 146, 28
- Cameron (1957) Cameron, A. G. W. 1957, PASP, 69, 201
- Carretta et al. (2002) Carretta, E., Gratton, R., Cohen, J. G., Beers, T. C., & Christlieb, N. 2002, AJ, 124, 481
- Casey (2014) Casey, A. R. 2014, ArXiv e-prints, arXiv:1405.5968
- Casey & Schlaufman (2015) Casey, A. R., & Schlaufman, K. C. 2015, ApJ, 809, 110
- Castelli & Kurucz (2004) Castelli, F., & Kurucz, R. L. 2004, ArXiv Astrophysics e-prints, astro-ph/0405087
- Cayrel et al. (2004) Cayrel, R., Depagne, E., Spite, M., et al. 2004, A&A, 416, 1117
- Cescutti et al. (2015) Cescutti, G., Romano, D., Matteucci, F., Chiappini, C., & Hirschi, R. 2015, A&A, 577, A139
- Christlieb et al. (2004a) Christlieb, N., Gustafsson, B., Korn, A. J., et al. 2004a, ApJ, 603, 708
- Christlieb et al. (2004b) Christlieb, N., Beers, T. C., Barklem, P. S., et al. 2004b, A&A, 428, 1027
- Cohen et al. (2013) Cohen, J. G., Christlieb, N., Thompson, I., et al. 2013, ApJ, 778, 56
- Cohen & Huang (2009) Cohen, J. G., & Huang, W. 2009, ApJ, 701, 1053
- Côté et al. (2017) Côté, B., Fryer, C. L., Belczynski, K., et al. 2017, ArXiv e-prints, arXiv:1710.05875
- Cowan et al. (2002) Cowan, J. J., Sneden, C., Burles, S., et al. 2002, ApJ, 572, 861
- Cowan et al. (2005) Cowan, J. J., Sneden, C., Beers, T. C., et al. 2005, ApJ, 627, 238
- Cui et al. (2013) Cui, W. Y., Sivarani, T., & Christlieb, N. 2013, A&A, 558, A36
- Dominik et al. (2012) Dominik, M., Belczynski, K., Fryer, C., et al. 2012, ApJ, 759, 52
- Drlica-Wagner et al. (2015) Drlica-Wagner, A., Bechtol, K., Rykoff, E. S., et al. 2015, ApJ, 813, 109
- Drout et al. (2017) Drout, M. R., Piro, A. L., Shappee, B. J., et al. 2017, ArXiv e-prints, arXiv:1710.05443
- Eichler et al. (2015) Eichler, M., Arcones, A., Kelic, A., et al. 2015, ApJ, 808, 30
- Ezzeddine et al. (2017) Ezzeddine, R., Frebel, A., & Plez, B. 2017, ApJ, 847, 142
- Farouqi et al. (2010) Farouqi, K., Kratz, K.-L., Pfeiffer, B., et al. 2010, ApJ, 712, 1359
- Fernández & Metzger (2013) Fernández, R., & Metzger, B. D. 2013, MNRAS, 435, 502
- François et al. (2016) François, P., Monaco, L., Bonifacio, P., et al. 2016, A&A, 588, A7
- Francois et al. (1993) Francois, P., Spite, M., & Spite, F. 1993, A&A, 274, 821
- Frebel & Bromm (2012) Frebel, A., & Bromm, V. 2012, ApJ, 759, 115
- Frebel et al. (2013) Frebel, A., Casey, A. R., Jacobson, H. R., & Yu, Q. 2013, ApJ, 769, 57
- Frebel et al. (2007a) Frebel, A., Christlieb, N., Norris, J. E., et al. 2007a, ApJ, 660, L117
- Frebel et al. (2007b) Frebel, A., Johnson, J. L., & Bromm, V. 2007b, MNRAS, 380, L40
- Frebel et al. (2010) Frebel, A., Kirby, E. N., & Simon, J. D. 2010, Nature, 464, 72
- Frebel & Norris (2015) Frebel, A., & Norris, J. E. 2015, ARA&A, 53, 631
- Frebel et al. (2016) Frebel, A., Norris, J. E., Gilmore, G., & Wyse, R. F. G. 2016, ArXiv e-prints, arXiv:1605.05732
- Frischknecht et al. (2016) Frischknecht, U., Hirschi, R., Pignatari, M., et al. 2016, MNRAS, 456, 1803
- Fulbright et al. (2004) Fulbright, J. P., Rich, R. M., & Castro, S. 2004, ApJ, 612, 447
- Geisler et al. (2005) Geisler, D., Smith, V. V., Wallerstein, G., Gonzalez, G., & Charbonnel, C. 2005, AJ, 129, 1428
- Gilmore et al. (2013) Gilmore, G., Norris, J. E., Monaco, L., et al. 2013, ApJ, 763, 61
- Goriely et al. (2011) Goriely, S., Bauswein, A., & Janka, H.-T. 2011, ApJ, 738, L32
- Goriely et al. (2015) Goriely, S., Bauswein, A., Just, O., Pllumbi, E., & Janka, H.-T. 2015, MNRAS, 452, 3894
- Gratton et al. (2000) Gratton, R. G., Sneden, C., Carretta, E., & Bragaglia, A. 2000, A&A, 354, 169
- Hansen et al. (2011) Hansen, C. J., Nordström, B., Bonifacio, P., et al. 2011, A&A, 527, A65
- Hansen et al. (2015) Hansen, T., Hansen, C. J., Christlieb, N., et al. 2015, ApJ, 807, 173
- Hansen et al. (2017) Hansen, T. T., Simon, J. D., Marshall, J. L., et al. 2017, ApJ, 838, 44
- Hayek et al. (2009) Hayek, W., Wiesendahl, U., Christlieb, N., et al. 2009, A&A, 504, 511
- Hill et al. (2017) Hill, V., Christlieb, N., Beers, T. C., et al. 2017, A&A, 607, A91
- Hill et al. (2002) Hill, V., Plez, B., Cayrel, R., et al. 2002, A&A, 387, 560
- Hirai et al. (2015) Hirai, Y., Ishimaru, Y., Saitoh, T. R., et al. 2015, ApJ, 814, 41
- Hirai et al. (2018) Hirai, Y., Saitoh, T. R., Ishimaru, Y., & Wanajo, S. 2018, ArXiv e-prints, arXiv:1801.08141
- Hollek et al. (2015) Hollek, J. K., Frebel, A., Placco, V. M., et al. 2015, ApJ, 814, 121
- Hollek et al. (2011) Hollek, J. K., Frebel, A., Roederer, I. U., et al. 2011, ApJ, 742, 54
- Honda et al. (2011) Honda, S., Aoki, W., Beers, T. C., & Takada-Hidai, M. 2011, ApJ, 730, 77
- Honda et al. (2007) Honda, S., Aoki, W., Ishimaru, Y., & Wanajo, S. 2007, ApJ, 666, 1189
- Honda et al. (2004) Honda, S., Aoki, W., Kajino, T., et al. 2004, ApJ, 607, 474
- Hotokezaka et al. (2015) Hotokezaka, K., Piran, T., & Paul, M. 2015, Nature Physics, 11, 1042
- Howes et al. (2015) Howes, L. M., Casey, A. R., Asplund, M., et al. 2015, Nature, 527, 484
- Howes et al. (2016) Howes, L. M., Asplund, M., Keller, S. C., et al. 2016, MNRAS, 460, 884
- Hunter (2007) Hunter, J. D. 2007, Computing in Science & Engineering, 9, 90. http://scitation.aip.org/content/aip/journal/cise/9/3/10.1109/MCSE.2007.55
- Ishigaki et al. (2010) Ishigaki, M., Chiba, M., & Aoki, W. 2010, PASJ, 62, 143
- Ishigaki et al. (2014) Ishigaki, M. N., Aoki, W., Arimoto, N., & Okamoto, S. 2014, A&A, 562, A146
- Ishigaki et al. (2013) Ishigaki, M. N., Aoki, W., & Chiba, M. 2013, ApJ, 771, 67
- Ishimaru et al. (2015) Ishimaru, Y., Wanajo, S., & Prantzos, N. 2015, ApJ, 804, L35
- Ivans et al. (2006) Ivans, I. I., Simmerer, J., Sneden, C., et al. 2006, ApJ, 645, 613
- Ivans et al. (2003) Ivans, I. I., Sneden, C., James, C. R., et al. 2003, ApJ, 592, 906
- Jacobson et al. (2015) Jacobson, H. R., Keller, S., Frebel, A., et al. 2015, ApJ, 807, 171
- Ji et al. (2015) Ji, A. P., Frebel, A., & Bromm, V. 2015, MNRAS, 454, 659
- Ji et al. (2016a) Ji, A. P., Frebel, A., Chiti, A., & Simon, J. D. 2016a, Nature, 531, 610
- Ji et al. (2016b) Ji, A. P., Frebel, A., Ezzeddine, R., & Casey, A. R. 2016b, ApJ, 832, L3
- Ji et al. (2016c) Ji, A. P., Frebel, A., Simon, J. D., & Chiti, A. 2016c, ApJ, 830, 93
- Johnson (2002) Johnson, J. A. 2002, ApJS, 139, 219
- Johnson & Bolte (2001) Johnson, J. A., & Bolte, M. 2001, ApJ, 554, 888
- Johnson & Bolte (2004) —. 2004, ApJ, 605, 462
- Jones et al. (2001) Jones, E., Oliphant, T., Peterson, P., et al. 2001, SciPy: Open source scientific tools for Python, , . http://www.scipy.org/
- Just et al. (2015) Just, O., Bauswein, A., Pulpillo, R. A., Goriely, S., & Janka, H.-T. 2015, MNRAS, 448, 541
- Kajino & Mathews (2017) Kajino, T., & Mathews, G. J. 2017, Reports on Progress in Physics, 80, 084901
- Kasliwal et al. (2017) Kasliwal, M. M., Nakar, E., Singer, L. P., et al. 2017, ArXiv e-prints, arXiv:1710.05436
- Kelson (2003) Kelson, D. D. 2003, PASP, 115, 688
- Kilpatrick et al. (2017) Kilpatrick, C. D., Foley, R. J., Kasen, D., et al. 2017, ArXiv e-prints, arXiv:1710.05434
- Kirby et al. (2008) Kirby, E. N., Simon, J. D., Geha, M., Guhathakurta, P., & Frebel, A. 2008, ApJ, 685, L43
- Koposov et al. (2015a) Koposov, S. E., Belokurov, V., Torrealba, G., & Evans, N. W. 2015a, ApJ, 805, 130
- Koposov et al. (2015b) Koposov, S. E., Casey, A. R., Belokurov, V., et al. 2015b, ApJ, 811, 62
- Korobkin et al. (2012) Korobkin, O., Rosswog, S., Arcones, A., & Winteler, C. 2012, MNRAS, 426, 1940
- Kratz et al. (2014) Kratz, K.-L., Farouqi, K., & Möller, P. 2014, ApJ, 792, 6
- Kratz et al. (2007) Kratz, K.-L., Farouqi, K., Pfeiffer, B., et al. 2007, ApJ, 662, 39
- Kupka et al. (1999) Kupka, F., Piskunov, N., Ryabchikova, T. A., Stempels, H. C., & Weiss, W. W. 1999, A&AS, 138, 119
- Kurucz (2011) Kurucz, R. L. 2011, Canadian Journal of Physics, 89, 417
- Lai et al. (2008) Lai, D. K., Bolte, M., Johnson, J. A., et al. 2008, ApJ, 681, 1524
- Lai et al. (2007) Lai, D. K., Johnson, J. A., Bolte, M., & Lucatello, S. 2007, ApJ, 667, 1185
- Lai et al. (2009) Lai, D. K., Rockosi, C. M., Bolte, M., et al. 2009, ApJ, 697, L63
- Lattimer et al. (1977) Lattimer, J. M., Mackie, F., Ravenhall, D. G., & Schramm, D. N. 1977, ApJ, 213, 225
- Lattimer & Schramm (1974) Lattimer, J. M., & Schramm, D. N. 1974, ApJ, 192, L145
- Lawler et al. (2015) Lawler, J. E., Sneden, C., & Cowan, J. J. 2015, ApJS, 220, 13
- Lawler et al. (2009) Lawler, J. E., Sneden, C., Cowan, J. J., Ivans, I. I., & Den Hartog, E. A. 2009, ApJS, 182, 51
- Li et al. (2015a) Li, H., Aoki, W., Zhao, G., et al. 2015a, PASJ, 67, 84
- Li et al. (2015b) Li, H.-N., Zhao, G., Christlieb, N., et al. 2015b, ApJ, 798, 110
- Lippuner & Roberts (2015) Lippuner, J., & Roberts, L. F. 2015, ApJ, 815, 82
- Margalit & Metzger (2017) Margalit, B., & Metzger, B. D. 2017, ApJ, 850, L19
- Marketin et al. (2016) Marketin, T., Huther, L., & Martínez-Pinedo, G. 2016, Phys. Rev. C, 93, 025805
- Mashonkina et al. (2010) Mashonkina, L., Christlieb, N., Barklem, P. S., et al. 2010, A&A, 516, A46
- Mashonkina et al. (2014) Mashonkina, L., Christlieb, N., & Eriksson, K. 2014, A&A, 569, A43
- Mashonkina et al. (2017) Mashonkina, L., Jablonka, P., Pakhomov, Y., Sitnova, T., & North, P. 2017, A&A, 604, A129
- Masseron et al. (2012) Masseron, T., Johnson, J. A., Lucatello, S., et al. 2012, ApJ, 751, 14
- Masseron et al. (2006) Masseron, T., van Eck, S., Famaey, B., et al. 2006, A&A, 455, 1059
- Masseron et al. (2014) Masseron, T., Plez, B., Van Eck, S., et al. 2014, A&A, 571, A47
- Mathews & Cowan (1990) Mathews, G. J., & Cowan, J. J. 1990, Nature, 345, 491
- Mckinney (2010) Mckinney, W. 2010
- McWilliam (1998) McWilliam, A. 1998, AJ, 115, 1640
- McWilliam et al. (1995) McWilliam, A., Preston, G. W., Sneden, C., & Searle, L. 1995, AJ, 109, 2757
- McWilliam et al. (2013) McWilliam, A., Wallerstein, G., & Mottini, M. 2013, ApJ, 778, 149
- Mendoza-Temis et al. (2015) Mendoza-Temis, J. d. J., Wu, M.-R., Langanke, K., et al. 2015, Phys. Rev. C, 92, 055805
- Metzger (2017) Metzger, B. D. 2017, ArXiv e-prints, arXiv:1710.05931
- Metzger et al. (2010) Metzger, B. D., Martínez-Pinedo, G., Darbha, S., et al. 2010, MNRAS, 406, 2650
- Meyer et al. (1992) Meyer, B. S., Mathews, G. J., Howard, W. M., Woosley, S. E., & Hoffman, R. D. 1992, ApJ, 399, 656
- Montes et al. (2007) Montes, F., Beers, T. C., Cowan, J., et al. 2007, ApJ, 671, 1685
- Morell et al. (1992) Morell, O., Kallander, D., & Butcher, H. R. 1992, A&A, 259, 543
- Mösta et al. (2017) Mösta, P., Roberts, L. F., Halevi, G., et al. 2017, ArXiv e-prints, arXiv:1712.09370
- Mumpower et al. (2016) Mumpower, M. R., McLaughlin, G. C., Surman, R., & Steiner, A. W. 2016, ApJ, 833, 282
- Nilsson et al. (2002) Nilsson, H., Zhang, Z. G., Lundberg, H., Johansson, S., & Nordström, B. 2002, A&A, 382, 368
- Nishimura et al. (2016) Nishimura, N., Podolyák, Z., Fang, D.-L., & Suzuki, T. 2016, Physics Letters B, 756, 273
- Nishimura et al. (2017) Nishimura, N., Sawai, H., Takiwaki, T., Yamada, S., & Thielemann, F.-K. 2017, ApJ, 836, L21
- Nishimura et al. (2015) Nishimura, N., Takiwaki, T., & Thielemann, F.-K. 2015, ApJ, 810, 109
- Nomoto et al. (2013) Nomoto, K., Kobayashi, C., & Tominaga, N. 2013, ARA&A, 51, 457
- Piro & Kollmeier (2017) Piro, A. L., & Kollmeier, J. A. 2017, ArXiv e-prints, arXiv:1710.05822
- Placco et al. (2014a) Placco, V. M., Frebel, A., Beers, T. C., et al. 2014a, ApJ, 781, 40
- Placco et al. (2013) —. 2013, ApJ, 770, 104
- Placco et al. (2014b) Placco, V. M., Frebel, A., Beers, T. C., & Stancliffe, R. J. 2014b, ApJ, 797, 21
- Placco et al. (2015) Placco, V. M., Beers, T. C., Ivans, I. I., et al. 2015, ApJ, 812, 109
- Placco et al. (2017) Placco, V. M., Holmbeck, E. M., Frebel, A., et al. 2017, ApJ, 844, 18
- Planck Collaboration et al. (2016) Planck Collaboration, Ade, P. A. R., Aghanim, N., et al. 2016, A&A, 594, A13
- Radice et al. (2016) Radice, D., Galeazzi, F., Lippuner, J., et al. 2016, MNRAS, arXiv:1601.02426
- Rembiasz et al. (2016) Rembiasz, T., Guilet, J., Obergaulinger, M., et al. 2016, MNRAS, 460, 3316
- Ren et al. (2012) Ren, J., Christlieb, N., & Zhao, G. 2012, A&A, 537, A118
- Rich & Boesgaard (2009) Rich, J. A., & Boesgaard, A. M. 2009, ApJ, 701, 1519
- Roberts et al. (2017) Roberts, L. F., Lippuner, J., Duez, M. D., et al. 2017, MNRAS, 464, 3907
- Roederer et al. (2014a) Roederer, I. U., Cowan, J. J., Preston, G. W., et al. 2014a, MNRAS, 445, 2970
- Roederer et al. (2008) Roederer, I. U., Lawler, J. E., Sneden, C., et al. 2008, ApJ, 675, 723
- Roederer et al. (2014b) Roederer, I. U., Preston, G. W., Thompson, I. B., Shectman, S. A., & Sneden, C. 2014b, ApJ, 784, 158
- Roederer et al. (2010) Roederer, I. U., Sneden, C., Thompson, I. B., Preston, G. W., & Shectman, S. A. 2010, ApJ, 711, 573
- Roederer et al. (2012) Roederer, I. U., Lawler, J. E., Sobeck, J. S., et al. 2012, ApJS, 203, 27
- Roederer et al. (2016) Roederer, I. U., Mateo, M., Bailey, III, J. I., et al. 2016, AJ, 151, 82
- Ryan et al. (1996) Ryan, S. G., Norris, J. E., & Beers, T. C. 1996, ApJ, 471, 254
- Ryan et al. (1991) Ryan, S. G., Norris, J. E., & Bessell, M. S. 1991, AJ, 102, 303
- Safarzadeh & Scannapieco (2017) Safarzadeh, M., & Scannapieco, E. 2017, MNRAS, 471, 2088
- Schatz et al. (2002) Schatz, H., Toenjes, R., Pfeiffer, B., et al. 2002, ApJ, 579, 626
- Schlafly & Finkbeiner (2011) Schlafly, E. F., & Finkbeiner, D. P. 2011, ApJ, 737, 103
- Shen et al. (2015) Shen, S., Cooke, R. J., Ramirez-Ruiz, E., et al. 2015, ApJ, 807, 115
- Shetrone et al. (2003) Shetrone, M., Venn, K. A., Tolstoy, E., et al. 2003, AJ, 125, 684
- Shetrone et al. (2001) Shetrone, M. D., Côté, P., & Sargent, W. L. W. 2001, ApJ, 548, 592
- Shibagaki et al. (2016) Shibagaki, S., Kajino, T., Mathews, G. J., et al. 2016, ApJ, 816, 79
- Simmerer et al. (2004) Simmerer, J., Sneden, C., Cowan, J. J., et al. 2004, ApJ, 617, 1091
- Simon et al. (2010) Simon, J. D., Frebel, A., McWilliam, A., Kirby, E. N., & Thompson, I. B. 2010, ApJ, 716, 446
- Simon et al. (2015) Simon, J. D., Drlica-Wagner, A., Li, T. S., et al. 2015, ApJ, 808, 95
- Sivarani et al. (2006) Sivarani, T., Beers, T. C., Bonifacio, P., et al. 2006, A&A, 459, 125
- Skrutskie et al. (2006) Skrutskie, M. F., Cutri, R. M., Stiening, R., et al. 2006, AJ, 131, 1163
- Skúladóttir et al. (2015) Skúladóttir, Á., Tolstoy, E., Salvadori, S., et al. 2015, A&A, 574, A129
- Sneden et al. (2008) Sneden, C., Cowan, J. J., & Gallino, R. 2008, ARA&A, 46, 241
- Sneden et al. (2016) Sneden, C., Cowan, J. J., Kobayashi, C., et al. 2016, ApJ, 817, 53
- Sneden et al. (2009) Sneden, C., Lawler, J. E., Cowan, J. J., Ivans, I. I., & Den Hartog, E. A. 2009, ApJS, 182, 80
- Sneden et al. (2014) Sneden, C., Lucatello, S., Ram, R. S., Brooke, J. S. A., & Bernath, P. 2014, ApJS, 214, 26
- Sneden et al. (1996) Sneden, C., McWilliam, A., Preston, G. W., et al. 1996, ApJ, 467, 819
- Sneden et al. (2003) Sneden, C., Cowan, J. J., Lawler, J. E., et al. 2003, ApJ, 591, 936
- Sneden (1973) Sneden, C. A. 1973, PhD thesis, The University of Texas at Austin.
- Sobeck et al. (2011) Sobeck, J. S., Kraft, R. P., Sneden, C., et al. 2011, AJ, 141, 175
- Spite et al. (2013) Spite, M., Caffau, E., Bonifacio, P., et al. 2013, A&A, 552, A107
- Spite et al. (2014) Spite, M., Spite, F., Bonifacio, P., et al. 2014, A&A, 571, A40
- Szentgyorgyi et al. (2016) Szentgyorgyi, A., Baldwin, D., Barnes, S., et al. 2016, in Proc. SPIE, Vol. 9908, Ground-based and Airborne Instrumentation for Astronomy VI, 990822
- Takiwaki et al. (2009) Takiwaki, T., Kotake, K., & Sato, K. 2009, ApJ, 691, 1360
- Taylor et al. (2014) Taylor, M., Cinabro, D., Dilday, B., et al. 2014, ApJ, 792, 135
- Tominaga (2009) Tominaga, N. 2009, ApJ, 690, 526
- Travaglio et al. (2004) Travaglio, C., Gallino, R., Arnone, E., et al. 2004, ApJ, 601, 864
- Tsujimoto & Shigeyama (2014) Tsujimoto, T., & Shigeyama, T. 2014, ApJ, 795, L18
- Umeda & Nomoto (2002) Umeda, H., & Nomoto, K. 2002, ApJ, 565, 385
- van der Walt et al. (2011) van der Walt, S., Colbert, S. C., & Varoquaux, G. 2011, Computing in Science & Engineering, 13, 22. http://scitation.aip.org/content/aip/journal/cise/13/2/10.1109/MCSE.2011.37
- Walker et al. (2015) Walker, M. G., Mateo, M., Olszewski, E. W., et al. 2015, ApJ, 808, 108
- Wallner et al. (2015) Wallner, A., Faestermann, T., Feige, J., et al. 2015, Nature Communications, 6, 5956
- Wanajo (2013) Wanajo, S. 2013, ApJ, 770, L22
- Wanajo et al. (2018) Wanajo, S., Müller, B., Janka, H.-T., & Heger, A. 2018, ApJ, 852, 40
- Wanajo et al. (2014) Wanajo, S., Sekiguchi, Y., Nishimura, N., et al. 2014, ApJ, 789, L39
- Waskom et al. (2016) Waskom, M., Botvinnik, O., drewokane, et al. 2016, seaborn: v0.7.0 (January 2016), , , doi:10.5281/zenodo.45133. http://dx.doi.org/10.5281/zenodo.45133
- Wehmeyer et al. (2015) Wehmeyer, B., Pignatari, M., & Thielemann, F.-K. 2015, MNRAS, 452, 1970
- Weisz et al. (2014) Weisz, D. R., Dolphin, A. E., Skillman, E. D., et al. 2014, ApJ, 789, 147
- Wenger et al. (2000) Wenger, M., Ochsenbein, F., Egret, D., et al. 2000, A&AS, 143, 9
- Westin et al. (2000) Westin, J., Sneden, C., Gustafsson, B., & Cowan, J. J. 2000, ApJ, 530, 783
- Westmeier et al. (2015) Westmeier, T., Staveley-Smith, L., Calabretta, M., et al. 2015, MNRAS, 453, 338
- Winteler et al. (2012) Winteler, C., Käppeli, R., Perego, A., et al. 2012, ApJ, 750, L22
- Woosley & Hoffman (1992) Woosley, S. E., & Hoffman, R. D. 1992, ApJ, 395, 202
- Wu et al. (2016) Wu, M.-R., Fernández, R., Martínez-Pinedo, G., & Metzger, B. D. 2016, MNRAS, 463, 2323
- Yong et al. (2013) Yong, D., Norris, J. E., Bessell, M. S., et al. 2013, ApJ, 762, 26
- Zhang et al. (2009) Zhang, L., Ishigaki, M., Aoki, W., Zhao, G., & Chiba, M. 2009, ApJ, 706, 1095
Appendix A References in halo and dwarf galaxy comparison sample
Allen et al. (2012); Aoki et al. (2002, 2005, 2007a, 2008, 2010); Barbuy et al. (2005); Barklem et al. (2005); Behara et al. (2010); Bonifacio et al. (2009, 2012); Burris et al. (2000); Caffau et al. (2011); Çalışkan et al. (2014); Carretta et al. (2002); Casey & Schlaufman (2015); Cayrel et al. (2004); Christlieb et al. (2004a); Cohen & Huang (2009); Cohen et al. (2013); Cowan et al. (2002); Cui et al. (2013); François et al. (2016); Frebel et al. (2007a, 2016); Fulbright et al. (2004); Geisler et al. (2005); Gilmore et al. (2013); Hansen et al. (2011, 2015); Hayek et al. (2009); Hollek et al. (2011, 2015); Honda et al. (2004, 2011); Howes et al. (2015, 2016); Ishigaki et al. (2010, 2013, 2014); Ivans et al. (2003, 2006); Jacobson et al. (2015); Johnson (2002); Johnson & Bolte (2004); Lai et al. (2007, 2008, 2009); Li et al. (2015b, a); Masseron et al. (2006); Mashonkina et al. (2010); Masseron et al. (2012); Mashonkina et al. (2014); McWilliam et al. (1995); Placco et al. (2013, 2014a, 2015); Rich & Boesgaard (2009); Roederer et al. (2010, 2014b); Ryan et al. (1991, 1996); Shetrone et al. (2001, 2003); Simon et al. (2010); Sivarani et al. (2006); Skúladóttir et al. (2015); Sneden et al. (2003); Spite et al. (2013, 2014); Westin et al. (2000); Yong et al. (2013); Zhang et al. (2009)