Stellar abundances and presolar grains trace the nucleosynthetic origin of molybdenum and ruthenium
aa r X i v : . [ a s t r o - ph . S R ] J un Astronomy & Astrophysics manuscript no. MoRucorfinallang c (cid:13)
ESO 2018October 22, 2018
Stellar abundances and presolar grains trace thenucleosynthetic origin of molybdenum and ruthenium
C. J. Hansen , A. C. Andersen , and N. Christlieb Landessternwarte, ZAH, Königstuhl 12, 69117 Heidelberg, Germany Dark Cosmology centre, Niels Bohr Institute, University of Copenhagen, Juliane Maries Vej 30, DK-2100 Copenhagen,Denmarktoday
ABSTRACT
This work presents a large consistent study of molybdenum (Mo) and ruthenium (Ru) abundances in the Milky Way.These two elements are important nucleosynthetic diagnostics. In our sample of 71 Galactic metal-poor field stars, wedetect Ru and/or Mo in 51 of these (59 including upper limits). The sample consists of high-resolution, high signal-to-noise spectra covering both dwarfs and giants from [Fe/H] = − . down to − . . Thus we provide information on thebehaviour of Mo I and Ru I at higher and lower metallicity than is currently known. In this sample we find a wide spreadin the Mo and Ru abundances, which is typical of heavy elements. We confirm earlier findings of Mo enhanced starsaround [Fe/H] = − . and add ∼ stars both dwarfs and giants with normal ( < . dex) Mo and Ru abundances, aswell as more than 15 Mo and Ru enhanced ( > . dex) stars to the currently known stellar sample. This indicates thatseveral formation processes, in addition to high entropy winds, can be responsible for the formation of elements likeMo and Ru. We trace the formation processes by comparing Mo and Ru to elements (Sr, Zr, Pd, Ag, Ba, and Eu) withknown formation processes. Based on how tight the two elements correlate with each other, we are able to distinguishif they share a common formation process and how important this contribution is to the element abundance. We findclear indications of contributions from several different formation processes, namely the p-process, and the slow (s-),and rapid (r-) neutron-capture processes. From these correlations we find that Mo is a highly convolved element thatreceives contributions from both the s-process and the p-process and less from the main and weak r-processes, whereasRu is mainly formed by the weak r-process as is silver. We also compare our absolute elemental stellar abundances torelative isotopic abundances of presolar grains extracted from meteorites. Their isotopic abundances can be directlylinked to the formation process (e.g. r-only isotopes) providing a unique comparison between observationally derivedabundances and the nuclear formation process. The comparison to abundances in presolar grains shows that the r-/s-process ratios from the presolar grains match the total elemental chemical composition derived from metal-poor halostars with [Fe/H] around − . to − . dex. This indicates that both grains and stars around and above [Fe/H] = − . are equally (well) mixed and therefore do not support a heterogeneous presolar nebula. An inhomogeneous interstellarmedium (ISM) should only be expected at lower metallicities. Our data, combined with the abundance ratios of presolargrains, could indicate that the AGB yields are less efficiently mixed into stars than into presolar grains. Finally, wedetect traces of s-process material at [Fe/H] = − . , indicating that this process is at work at this and probably at evenlower metallicity. Key words.
Stars: abundances, Stars: general, meteorites, Galaxy: evolution, Galaxy: solar neighbourhood
1. Introduction
Nucleosynthetic processes are fundamental to the exis-tence of stars, planets, and life. The neutron-capture pro-cesses can be traced through abundances of heavy elements(Z > ). Molybdenum and ruthenium are excellent traceelements because over time they can probe various enrich-ment scenarios, that contribute to the chemical evolutionof our Galaxy in stars, meteorites, and planets. These twoelements are created by three different processes, namelya rapid and slow neutron-capture process (the r- and s-processes, respectively), and a p-process.Owing to the lack of hyperfine structure, isotopic ratiosof Mo and Ru cannot be determined by the stellar spec-tra covering a near-UV to visual range, and we can there- Send offprint requests to : cjhansen, e-mail: @lsw.uni-heidelberg.de fore only derive elemental stellar abundances. However, el-emental abundances can be used to trace major contribu-tions from various nucleosynthetic processes (as shown inFrançois et al. 2007; Roederer et al. 2010a; Hansen et al.2012). The consistently analysed large sample allows us tomake an indepth comparison of our data to abundancesmeasured in presolar grains. The grain abundances directlyprobe which of the three processes — r-, s-, or p — con-tributed most to the stellar gas, as well as to the presolargrains at a given [Fe/H]. Furthermore, we can test whetherthere are changes as a function of time or metallicity in thiscontribution or, when phrased differently, if the dominatingformation process varies. The formation processes that wetry to trace may be associated with various sites, and herewe only list a few (for details we refer to Fröhlich et al. inprep.).
Article number, page 1 of 18 &A proofs: manuscript no. MoRucorfinallang
The r-process seems to split into two, a main and aweak channel, where the main r-process could be linkedto neutron-star mergers (e.g. Freiburghaus et al. 1999;Goriely et al. 2011), while the weak r-process may comefrom neutrino-driven winds (e.g. Arcones & Montes 2011),or electron-capturing supernovae (SNe) collapsing on O-Mg-Ne cores (Wanajo et al. 2011). A p-process contribu-tion could also be expected from supernova winds thatwould facilitate for example a ν p-process (Fröhlich et al.2006). Finally, the parameter study of Farouqi et al. (2009)indicates that various r-process elements can be createdin neutron-rich high entropy winds (HEW). Later in theGalactic chemical evolution s-process contributions fromthe main s-process associated with low-mass asymptotic gi-ant branch (AGB) stars (Bisterzo et al. 2011; Karakas et al.2012; Stancliffe et al. 2011) and the weak s-process from,for instance, massive fast-rotating stars (see Pignatari et al.2010; Frischknecht et al. 2012) are expected to contributeto and possibly dominate the gas composition later on .For reviews covering several of these formation processes,see Sneden et al. (2008) and Käppeler et al. (2011).Very few studies of both Ru and Mo abundances inmetal-poor field stars have been performed. This is in partbecause their main absorption lines lie in the blue or near-UV part of the spectra, requiring very high-quality bluespectra for these studies. Currently, our results composethe largest, most consistent sample of stars, for which Moand Ru have been derived using 1D, local thermodynamicequilibrium based codes. Peterson (2011) performed a largestudy of neutral and ionised Mo and Ru in both the near-and UV spectral ranges of a sample centered on [Fe/H] ∼ − . . In our sample the stars span a slightly broadermetallicity range and a wider Mo and Ru abundance rangethan previous studies. Other studies have derived Mo I forsingle r-process enhanced stars (e.g. Sneden et al. 2003;Ivans et al. 2006; Honda et al. 2007; Mashonkina et al. 2010;Roederer et al. 2010b; Siqueira Mello et al. 2013). Roedereret al. (2012) also detected Mo II and Ru II in four ex-tremely metal-poor stars from UV spectra. The sparse Modetections of metal-poor halo field stars in the literatureare also confirmed by Fröhlich et al. (in prep), who carriedout an extensive literature study. Furthermore, detectionsof Mo in globular clusters have been made by Yong et al.(2008), Lai et al. (2011), and Roederer et al. (2011), wherethe latter also report Ru abundances. Ruthenium was alsostudied in a metal-rich sample mainly consisting of Ba-stars(Allen & Porto de Mello 2007). Here we focus on metal-poorfield stars and not on stars that may have been polluted bya binary or cluster environment.Isotopic abundance ratios can be derived from presolargrains and other meteoritic samples. We present, for thefirst time, a comparison of Mo abundances in grains andstars spanning a broad range of metallicities. While mostof the material that went into making the solar system wasthoroughly processed and mixed, thus losing isotopic het-erogeneity and all memory of its origin, small quantities of refractory presolar dust grains have been proven to bestardust with their original stellar isotopic signatures in-tact (Hoppe & Zinner 2000; Nittler 2003; Zinner 2005, 1998; According to Cescutti et al. (2013), these objects may alsocontribute with ‘primary’ weak s-process yields very early inthe history of our Galaxy. Ott & Hoppe 2007; Alexander et al. 2007; Lodders 2010).These reflect the nucleosynthetic fingerprint of their stellarproduction site.The first presolar grains in meteorites were isolated byLewis et al. (1987) and identified as tiny diamonds. Latersilicon carbide (SiC; Bernatowicz et al. 1987), graphite(Amari et al. 1990), corundum (Al O ; Hutcheon et al.1994; Nittler et al. 1994), silicon nitride (Si N ; Russellet al. 1995; Nittler et al. 1995), and spinel (MgAl O ;Nittler et al. 1994) were identified. Some of the SiC andgraphite grains have been found to carry small inclusionsof Ti-, Mo-, and Zr-carbides (Bernatowicz et al. 1991). Cen-tral to the identification of presolar grains is to determinethe isotopic composition of the grain and/or some trace el-ements trapped in the grains. As a rule the isotopic compo-sition of the grain or some of its inclusions deviates stronglyfrom the normal solar system composition. The isotopic sig-natures of the grains contain information about the nucle-osynthesis processes of the parent stars. Information on in-dividual stars can be obtained by studying single grains by,for example, SIMS (secondary ion mass spectrometry) forthe light-to-intermediate-mass elements, RIMS (resonanceionization mass spectrometry) for the heavy elements, andlaser heating and gas mass spectrometry for He and Ne.TEM (transmission electron microscopy) and SEM (scan-ning electron microscopy) is used to study the crystal struc-ture of the individual grains.The best characterised of all the presolar grains areSiC (6 ppm) grains, and almost all of them originate inAGB stars. Graphite (less than 1 ppm) was traditionallyassumed to originate in supernovae, but more detailed mea-surements of s -process signatures indicate that most high-density graphite grains are more likely to originate in AGBstars (Croat et al. 2005). Since presolar grains seem to comefrom many stellar sources and to be relatively unprocessed,they also provide constraints on circumstellar dust produc-tion rates in the Galaxy. The relative abundances of thedifferent types of presolar grains seem to indicate that AGBstars are the main dust producers in the Galaxy, while su-pernovae only contribute a few percent (Alexander et al.2007).Isotopic abundances of Mo and Ru have been measuredin presolar SiC and graphite grains. Since the amounts arequite small within the individual grains, it is on the edge ofwhat is possible with the current techniques. In works byBecker & Walker (2003), Lee & Halliday (2002), and Halli-day (2003) it seems that the Mo isotopic abundances behavenormal and seem to be well mixed into the matrix materialof the meteorites, whereas Yin et al. (2002); Dauphas et al.(2004); Chen et al. (2004), on the other hand, find abun-dance anomalies for Mo (and Ru), indicating that the solarnebula was not homogeneous when the meteorites and theEarth formed.In this paper, we aim at finding evidence of any of theseconclusions by studying elemental abundance in a large stel-lar sample and compare these to isotopic SiC grain abun-dances.Here we focus on the sample and stellar parameters inSect. 2, and present the analysis in Sect. 3. The results willbe described in Sect. 4, along with detailed abundance cor-relations from which clues to the nucleosynthetic formation neither when the core formed nor later when the mantleformedArticle number, page 2 of 18ansen, C. J. et al.: Molybdenum and ruthenium processes that enriched the extremely metal-poor stars canbe drawn. Finally, a detailed comparison of meteoritic iso-topic abundances and elemental stellar abundances can befound in Sect. 5 along with our discussion, which will befollowed by a short summary and conclusion in Sect. 6.
2. Data and stellar parameters
This study is based on the analysis of the stellar sam-ple described in Hansen et al. (2012). The sample con-sists of 71 dwarfs and giants, observed with high-resolutionspectrographs (UVES/VLT - Dekker et al. (2000), andHIRES/Keck - Vogt et al. (1994)). The spectra are of R ∼ − ∼ in the blue. Owing to the ex-cellent spectrum quality (a typical signal-to-noise ratio isS/N > at 3200Å ), abundances of Mo I and Ru I couldbe derived from blue and near-UV spectral lines. Details ondata reduction can also be found in Hansen et al. (2012). Continuum placement is a crucial part of the data analy-sis, especially when working in the blue spectral range. Asnoted in Peterson (e.g. 2011, 2013), an over- or underes-timation of the continuum level will lead to over- or un-derestimated Ru (or Mo) abundances. Therefore, we testedseveral fitting functions to ensure an accurate continuumplacement. The echelle spectra were normalised in IRAF us-ing the ‘continuum’ package. We explored four fitting func-tions by varying the order of the fitted polynomial betweenone and seven and selecting the optimal fitting function,which may vary depending on the temperature, gravity, andmetallicity of the stars. We found that the best functionswere either fourth-order cubic splines or sixth-order Legen-dre polynomials. The function is fitted to the observed spec-tra, rejecting points that lie more than one sigma above orbelow the fitted function, and this process is repeated itera-tively up to 30 times until the best match is found. Then theobserved spectrum is divided by the best fit function (in-cluding rejected points). This yields well-normalised spec-tra overall; however, each spectrum was visually inspected,and the procedure was repeated with different functions incase the result using for example a cubic spline was notsatisfying. Before the abundances were derived, a region of ± Å around the line of interest was zoomed in on and thecontinuum was re-evaluated and manually optimised, usingthe synthetic spectrum code MOOG (Sneden 1973, version2010). By choosing a range of 40 Å, we avoid local biasesthat can be introduced by molecular bands, if for instancea region of only ± Å was selected. However, with the 40 Åregion we optimise the continuum, since this ensures thatwe include both absorption features and continuum pieces.Furthermore, it should be noted that the stellar samplewas composed of stars without carbon or nitrogen enhance-ments. This means that there will be no strong molecularbands interfering with the continuum placement.To estimate how much this approach to setting the con-tinuum level affects the derived abundances, every abun-dance was derived at least twice with a period of a cou-ple of weeks between each abundance derivation. In thisway, the derivations are largely independent, and the con-tinuum placement, treatment of blends, and the overall fit-ting procedure can be assessed. For the most extreme cases, that is stars with high metallicity or metal-poor stars withslightly noisier spectra, the abundances derived in each ofthe analysis runs would differ by ∼ . dex. In the spectraof the highest quality (high resolution and S/N), the abun-dances of each of the elements would typically agree within0.01 to 0.03 dex. Thus, the uncertainties related to fittingsynthetic spectra spans a range of 0.01–0.1 dex in the de-rived abundances (for details on abundance uncertaintiessee Sect. 3.3). Since the blue region is a crowded spectral region, lineblending may be a concern, and it is therefore importantto have a very complete description of lines in the vicinityof the line of interest. Thus, the line list was optimised asdescribed below in Sect. 3. With this list at hand we canmimic the wing blends on each side of the Ru line, as wellas reproduce the observed spectrum around Mo (see Figs.1 and 2). As mentioned above, no star with enhanced levelsof N or C have been included in the sample. The effect ofmolecules (e.g. CN, CH, NH, etc.) was tested and found tobe of the order of -0.02 dex for Mo and -0.01 dex for Ru. Thetest was carried out on the cold giant BD+01 2916 wherethe effect of molecules is expected to be stronger than inwarmer stars. From this the continuum placement is seen tobe more important, influencing the Mo and Ru abundancesmore than molecular blends (owing to the sample selection).This means that molecular features are very weak, and theline blending is therefore reduced to atomic lines.Finally, the broadening of the synthetic spectra was setin accordance with the instrumental resolution and fine-tuned so that the lines of moderate strength in the blueregion would be matched. This was done by adjusting thefull width half maximum in MOOG, and it was kept fixedwhen synthesising spectral lines.
The stellar parameters are taken from Hansen et al. (2012)and listed in Tables 4 and 5. For the vast majority of thestars, these parameters have been determined using the fol-lowing methods. The effective temperatures are estimatedusing V − K-photometry (in a few cases we had to use ex-citation potential owing to the lack of K-magnitudes). Thegravities are based on parallaxes when these are available(otherwise the ionisation equilibrium of Fe I and Fe II hasbeen enforced), and the mictroturbulence is determined byrequiring that all Fe I lines give the same abundance regard-less of line strength (equivalent width). Finally, the metal-licity ([Fe/H]) was derived from an average of Fe I and IIabundances since these agree in general. For a few stars inwhich there is a large difference between Fe I and II, wehave chosen the Fe II in order to minimise NLTE effects onthe [Fe/H]. However, in some of the extremely metal-poorstars very few or no Fe II lines were present in spectra,and for these [Fe/H] has been based on Fe I abundance.For each set of stellar parameters (temperature, gravity,and metallicity) a MARCS model (Gustafsson et al. 2008)was computed using the interpolation code from (Masseron2006). If the temperatures are derived from Fe lines, or ifthe gravity is set by enforcing ionisation equilibrium, theparameters have been labelled with an ‘a’ in Tables 4 and
Article number, page 3 of 18 &A proofs: manuscript no. MoRucorfinallang
Table 1.
Atomic data from VALD and solar abundances fromAnders & Grevesse (1989).
Element λ χ log gf log ǫ ⊙ [ Å ] [ eV ] [ dex ] Ru I 3498.94 0.00 0.31 1.84Mo I 3864.10 0.00 -0.01 1.92Mo I 5506.49 1.34 0.06 1.925. In some cases the photometry or de-reddening were un-certain, while in others the parallaxes were inadequatelydetermined, and the stellar parameters are tagged with a‘c’ to indicate they have been adjusted via excitation poten-tials and/or imposing ionisation equilibrium. Finally, starswith a special r-process pattern have been highlighted witha ‘b’ in the same tables.In general these stellar parameters agree with those re-cently derived for the Gaia-ESO benchmark stars (Jofreet al. 2013). For two of the three stars we have in com-mon with their sample (HD140283 and HD22879) the tem-peratures agree within 18 K, the gravities within 0.07 dex,and the [Fe/H] within 0.17 dex. We also have HD122563in common with their work, but here the differences arelarger (48 K, 0.7 dex, and 0.22 dex). However, when we com-pare our stellar parameters to those derived by Hondaet al. (2004, 2007), we find good agreement (18 K, 0.2 dex,0.2 dex) for HD122563 and HD88609.
3. Abundance analysis
The abundances were derived based on a line list down-loaded from VALD (Kupka F. 2000) using ‘extract all’. Thelines of interest as well as blends were cross checked withvalues from NIST (Kramida et al. 2013) and similar stud-ies from the literature. The log gf values we adopted forMo I and Ru I (Table 1) from VALD are the same as thosegiven in Sneden et al. (2003); Ivans et al. (2006); Peter-son (2013), and NIST. Additional line list information onmolecules was taken from the database of Kurucz.The lines we focussed on for this study are Mo I at3864.1 Å and Ru I at 3498.9 Å. A lot of effort is currentlygoing into assembling line lists for large surveys such as theGaia-ESO Survey (GES). Thus we list the details on lineswe have included and checked visibility in the solar spectrain Table 2. Especially the Mo I line at 5506Å has beenchecked since it is included in the GES line list.Out of these fairly strong (persistent ) lines (see Table2) only five Mo I and five Ru I lines (‘*’) were detectable.The bluest Mo I line is very close to the core of a hydrogenline and is therefore discarded, the three reddest Mo I linesare very weak, and the red 5506 Å line becomes too weakto detect below [Fe/H] = − in dwarfs. The red line willtherefore only be used in the few ‘metal-rich’ dwarf stars National Institute of Standards and Technology –http://physics.nist.gov/asd For example physics.nist.gov/PhysRefData/Handbook/Tables/rutheniumtable3.htm
Table 2.
All persistent lines in the NIST database of Mo andRu. Visible lines in the high-resolution solar NOAO atlas aremarked with ‘*’.
Mo I Ru I3133.59 Å 3437.74 Å *3158.17 Å 3498.94 Å *3170.34 Å 3589.22 Å3193.98 Å 3593.03 Å3208.84 Å 3596.19 Å3447.12 Å 3726.93 Å3798.25 Å * 3728.03 Å *3864.10 Å * 3730.43 Å3902.95 Å 3798.90 Å *4069.88 Å 3799.35 Å *4188.32 Å 4199.89 Å4411.70 Å ...5506.49 Å * ...5533.03 Å * ...5570.44 Å * ...
Fig. 1.
Synthetic spectrum fit to the red Mo line at 5506 Å inthe metal-rich dwarf HD113679 ([Fe/H] = − . ). Spectra havebeen calculated with [Mo/Fe] = − ., . , . , and 0.45. (see Fig. 1) in the sample . This leaves us with one con-sistently useful Mo line, namely the 3864 Å Mo I line. Theblending Fe lines were updated using atomic data from Ku-rucz’s data base , since data neither from NIST nor VALDcould properly reproduce the strong Fe-Sc-Cr blend blue ofthe 3864 Å Mo I line.Similar findings are made for the Ru I lines. Here thetwo reddest lines fall in a strong hydrogen line, the 3728 Åline is heavily blended, and the blue most 3437 Å is blendedand located in a NH band. The 3499 Å is the strongest andcleanest of all the detectable Ru lines. The 3864 Å Mo line isblended (but it is the only usable line), while the Ru 3499 Å As seen from Table 5, the red 5506 Å line generally yieldsabundances for stars with [Fe/H] > − . dex unless the star isenhanced in Mo. Fig. 2.
Synthetic spectrum fit to the Mo line at 3864Å (top) andthe Ru line 3498 Å (bottom) in the metal-poor dwarf HD298986.The signal-to-noise ratio is a little lower in this spectrum com-pared to that in Fig. 3. Spectra have been calculated with[X/Fe]= -5., 0.5 ± . or 0.66 ± . , where X is Mo or Ru, re-spectively. line is fairly strong and clean (with only a weak blend inthe red wing – see Fig. 2). Therefore, this line yields themost trustworthy abundances. Further information can befound in Table 1. Additionally, by only relying on these twolines, we can directly compare our derived abundances tothose presented in Peterson (2013), which is currently theonly other study detecting Mo and Ru in more than tenfield stars in the Galaxy. Figure 2 shows one of the mostmetal-poor warm dwarfs (HD298986 with [Fe/H] = − . ),for which both Mo and Ru has been derived to date .For comparison we have included a figure of HD189558that has an even higher signal-to-noise ratio than that ofHD298986. A bit of noise < is seen in the Mo line,and slightly more around the Ru line in the spectrum ofHD298986. In HD189558 an almost perfect match betweenobserved and synthesised spectrum is seen (Fig 3). Thespectra of both stars have been obtained with a resolutionof ∼ , which for comparison is higher than the typicalspectra ( R ∼ ) presented in Peterson (2013). HD188510 is 0.1 dex more metal-poor, but the spectrum is ofslightly lower quality.
Fig. 3.
High signal-to-noise spectrum of HD189558 with syn-thetic spectra of [Mo/Fe] = − , . , . , . dex (top), and[Ru/Fe] = − , . , . , . dex (bottom) over-plotted. The stellar abundances were derived using MARCS 1Dmodels (Gustafsson et al. 2008), and MOOG (Sneden 1973)LTE synthetic spectrum code (version 2010). Our sampleconsists of 42 dwarfs and 29 giants, in which we detecteither Mo and/or Ru in 30 (33) dwarfs and 22 (26) gi-ants (including upper limits). We derive abundances onlyfrom neutral Mo and Ru lines. Both Mo and Ru have beendetected down to [Fe/H] = − . in HD115444, for whichWestin et al. (2000) did not have the needed spectral cov-erage to determine Ru. Upper limits for Mo and Ru arefound for one of our most metal-poor stars (HD126587)with [Fe/H] = − . . In this sample we have 12 starsin common with the literature, namely CS31082-001 forwhich Siqueira Mello et al. (2013) derived [Mo/Fe] = 0 . and [Ru/Fe] = 1 . , which is in excellent agreement withour values of 0.94 and 1.44, respectively. The stellar pa-rameters are also in good agreement (well within the com-bined errors), except for the microturbulence, which doesnot have any (or a very little) impact on these abundancessince most of these are derived from weak lines. We alsostudied the two stars, HD88609 and HD122563, for whichwe find a good agreement with the stellar parameters pre-sented in Honda et al. (2007), as mentioned in Sect. 2.3. Article number, page 5 of 18 &A proofs: manuscript no. MoRucorfinallang
We find a maximum difference of 0.19 dex between ourMo abundances and their results, a value that is withinthe combined uncertainties. Three of the dwarf stars areHD106038, HD160617, and HD188510, which we have incommon with Peterson (2013). Our derived abundances areeither in perfect agreement with those derived in Peterson(2013) or agree within 0.1 dex. The 0.1 dex difference in Mowas found for HD106038, for which we derived 0.18 dex dif-ferent metallicities. We have two stars in common with Pe-terson (2011), namely HD76932, for which we find an excel-lent agreement between the stellar parameter derived, andthe difference of 0.15 dex in [Mo/Fe] probably stems fromcontinuum placement. The second object is HD140283, forwhich neither we nor Peterson (2011) could detect any MoI nor Ru I.However, Peterson (2011) studied UV spectra around λ = 2000 Å and could therefore derive abundances of MoII and place an upper limit on Ru II in HD140283. Sincethe line lists for the UV region are often incomplete, owingto the lack of atomic data, it is common practice to em-pirically add oscillator strengths or even artificial lines toachieve a better match between the observed UV spectraand the synthesised ones. In Peterson (2011), they increasethe log gf value of Mo II by 0.133 dex for five Mo II linesin the UV, to scale their theoretical values to that of themeasured log gf for the Mo II line at 2082 Å. They notethat the log gf values of Mo I and Mo II ‘appear to be on aconsistent scale’ . Another study of Mo II in the UV is thatof Roederer et al. (2012), in which the 2871.51 Å line wasused. In this study, they find a difference of up to ∼ . dexbetween the Mo I abundances and the Mo II upper lim-its. Since the studies in the UV often use adjusted log gfvalues, we would like to stress that possible NLTE effectsmight still affect both the Mo and Fe abundances, and thatthese effects are often seen to increase with increasing linestrength (Lind et al. for Fe and for Sr see 2012 Andrievskyet al. 2011; Bergemann et al. 2012 or Hansen et al. 2013).A larger study of Mo and Ru is that of Peterson (2013),in which Mo is detected in 20 stars (28 including upper lim-its). In this study the stellar parameters were determinedfrom spectra (as ours labelled ‘a’ or ‘c’ in Tables 4 and5), and such stellar parameters can be off due to NLTEeffects on Fe especially in stars with higher temperatures,lower metallicities, and gravities (Lind et al. 2012). TheNLTE corrections to the stellar parameters may in turnpropagate into the stellar abundances and cause a largerchange in abundance than the actual NLTE correction tothe abundance itself. (This was shown to be the case for SrII Hansen et al. 2013.) Peterson (2013) presents Mo I andRu I abundances (between 0.4 – 0.8 dex). The two elementsagree within 0.1 dex, and show a low star-to-star abundancescatter. Therefore, they average the abundances of Mo andRu and present this single value in order to better discussthe abundance enhancement with respect to other elements.Our results deviate from this picture (see Fig. 4, as wellas Tables 4, 5) because we find abundances in this largersample that are not only enhanced but also only slightlyabove solar (also found by Honda et al. 2007; Roedereret al. 2012); in other words, two groups with different en-hancements of Mo and Ru are found. Here we find a largestar-to-star abundance scatter as for most other heavy el- For comparison to Sr I and II we refer to Hansen et al. (2013)and Sect. 4 ements (e.g. François et al. 2007; Roederer et al. 2010a;Hansen et al. 2012). Furthermore, we show that Mo and Ruin several stars deviate by more than 0.1 dex and that theelements are formed via different channels. Averaging theseabundances therefore seems to erase important information,and we therefore choose to keep all heavy element abun-dances apart, compare them on a log ǫ abundance scale,and discuss their behaviour individually.We derive Mo and Ru abundances that are only en-hanced with a factor of two to three compared to the sur-rounding elements Zr, Pd, and Ag. However, we also findstars in the sample with Mo and Ru abundances that are ofsimilar size as the Zr, Pd, and Ag abundances from Hansenet al. (2012). Therefore, other sites than the high entropywinds (HEW; Farouqi et al. 2009) may be equally feasible(see Fröhlich et al. 2014 in prep. for further comments onmodels and possible sites). There is also an overlap betweenthis study and the four giants in Roederer et al. (2012).Most of the stellar parameters are in good agreement withthose presented here (adopted from Hansen et al. 2012), andthe derived Mo and Ru abundances agree within 0.1 dex. Aslight disagreement in the determined stellar parameters ofHD108317 causes a slightly larger difference in the derivedabundances of up to 0.24 dex. This difference is, however,well within the combined abundance uncertainties.To obtain the purest trace of the nucleosynthetic for-mation processes, absolute (log ǫ ) abundances have beencompared to those from Hansen et al. (2012). The abun-dances have been obtained in exactly the same way us-ing the same atmospheric models, methods for determiningstellar parameters, as well as the same synthetic spectrumcode (though here we used a more recent version). Thisallows a direct and unbiased comparison to those abun-dances (of Sr, Zr, Pd, Ag, Ba, and Eu). The derived Moand Ru abundances have been plotted vs [Fe/H] in Fig. 4.The [Mo/Fe] and [Ru/Fe] have been calculated using theMo ⊙ = 1 . and Ru ⊙ = 1 . from Anders & Grevesse(1989, see our Table 1), which will facilitate a direct com-parison to the meteoritic abundances discussed in Sect. 5. The abundance uncertainties resulting from uncertaintiesin stellar parameters, continuum placement in the near-UV, and fitting technique have been determined in greatdetail for a sample dwarf and giant star. The two se-lected stars are listed in Table 3. They were chosensince they have a good, clean spectral region aroundthe lines. This allows an easier assessment of the stel-lar parameter’s impact on the abundances. Furthermore,these stars have stellar parameters that are representa-tive of the dwarfs and giants in this sample. The gen-eral uncertainties on the stellar parameters adopted fromHansen et al. (2013) are for temperature/gravity/[Fe/H]/ ξ : ±
100 K / . / . / . kms − , and fitting related errorsspan a factor of 10 when associated with fitting technique(best case ± . dex) and continuum placement ( ± . dex).The uncertainty related to the synthetic spectrum fittingand continuum placement also includes an assessment ofthe line blending. The strength of the blending lines werevaried to test the impact on the derived abundances. Thetotal propagated error is given in Table 3, where each termhas been added in quadrature. Article number, page 6 of 18ansen, C. J. et al.: Molybdenum and ruthenium
Table 3.
Abundance uncertainties given on a log ǫ − scale for a dwarf (HD188510) and giant (BD+54 1323) star. Dwarf GiantParameter Mo Ru Mo RuT ± K 0.11 0.1 0.08 0.13 log g ± . [ Fe / H ] ± . ξ ± . ± : 0.11(0.15) 0.12(0.15) 0.11(0.14) 0.14(0.16)As the continuum in the near-UV is often challengingto place, particularly in stars with [Fe/H] > − . , we adoptthe larger uncertainty given in parentheses in all the figures.The uncertainty is ∼ . dex on average.Generally, the spectra could be well fitted; however, ina few cases where the spectral quality was a bit lower, orwhere the metallicity of the star was high, the line fit wouldbecome too poor, and we could therefore only place up-per limits. More than four observational data points hadto be well matched for the fit to be a detection, otherwiseit would lead to upper limits. We have a few such cases inour sample, where the metallicity is above − dex (caus-ing line blending), and the spectral quality is lower, namelyHD175179, G005-040, HD105004, CD-45 3283. The two for-mer stars also have upper limits on Sr and Y as shown inHansen et al. (2012). The molybdenum falls at the end ofthe blue spectrum (when using the UVES/VLT 346nm set-ting), and this means that the spectra generally will bemore noisy around the Mo line than around the Ru line.For the last two stars, as well as HD111980, the spectraend less than 2-3Å from the Mo line, either due to radialvelocity and/or spectrum quality, and we can therefore notplace the continuum properly in this noisy region. Thus wepresent upper limits for these stars.
4. Results
In total we derived Mo abundances for 47 stars and Ruabundances for 58 stars, with clear detections down to − . in [Fe/H]. The results are listed in Tables 4 and 5 and theabundances relative to iron are shown in Fig. 4. As for manyof the heavy elements, Mo and Ru also show a large star-to-star scatter. For the Mo and Ru enhanced stars, we finda good agreement and overlap with the metal-poor samplefrom Peterson (2013).Figure 4 shows our [Mo/Fe] and [Ru/Fe] compared toother samples as a function of [Fe/H]. The comparison sam-ples are comprised of Peterson (2013) study, and all theother studies of less than ten stars have been grouped to-gether. Here we note that only stars that were labelled asnot being Ba stars from the sample of Allen & Porto deMello (2007) have been included. This is a large study of39 stars (33 Ba stars) for which normal (extreme) Ru abun-dances have been derived for stars with metallicities above ∼ − . dex. Peterson (2011) is incorporated in the ‘Others’sample, as is the extremely Ru enhanced bulge giant starfrom Johnson et al. (2013), and the upper limits on Mofrom Roederer et al. (2014). For the 12 stars we have incommon with the literature we have shown both our ownand previous measurements (using different symbols), andfor eight of these duplicates the agreement is so good that the symbols touch or overlap. For the remaining stars adifference of ∼ . dex in either Mo or Fe lead to largerabundance differences (see Sects. 2 and 3.2). Stars withnormal and enhanced Ru and Mo abundances are foundboth in dwarfs and giants, regardless of which method wasapplied to determine stellar parameters. We note that theexact abundances of dwarfs and giants could change due toNLTE effects on stellar parameters, as well as on Mo and Ruabundances . There is a slight gap around [Fe/H] = − . in [Mo/Fe]. More Mo detections around this [Fe/H] wouldbe needed to see whether the enhanced trend in the dwarfswould continue towards lower metallicities, or if they wouldfollow the chemical normal pattern. However, this could bean observational bias that is not easily overcome (see be-low).Since the stellar abundances may be biased by the 1D,LTE assumptions, we tried to assess the bias this assump-tion may lead to by separating dwarfs and giants, since theymost likely will be affected differently by these assumptions.Thus, an offset in trend found between the dwarfs and gi-ants can be used as a possible probe of the size and impactof these assumptions compared to 3D, NLTE abundances.Unfortunately, no model atom exits for Mo or Ru, which iswhy we cannot correct our LTE abundances.A difference in the abundance behaviour between dwarfsand giants can also be induced by a difference in metallicity.Most of the Mo and Ru absorption lines disappear in the ex-tremely metal-poor dwarfs, while they remain detectable inthe giants. This means that the giants can probe the abun-dance behaviour at lower metallicities, and will therefore(despite possible deviations from LTE and 1D) be bettertracers of the earliest formation processes (r- and possiblyp-processes). The dwarfs, on the other hand, mainly showMo and Ru lines above [Fe/H] ∼ − , and will therefore tracelater epochs of the chemical evolution of the Galaxy, whichhave been enriched (and dominated?) by the s-process(es)and smaller p-process contributions (see Fig. 4). At highmetallicities ([Fe/H] > − . ), we can expect to see both s-and p-process contributions from SN type Ia. The exactyields are model sensitive (see Travaglio et al. 2011, for de-tails). This will be the overall trend we can expect to seefrom the elemental abundances, whereas the isotopic abun-dances of Mo and Ru will each have a direct link to theindividual formation process(es) that create that particularisotope. For guidance these have been listed in Table 6. A similar study was carried out by Hansen et al. (2013) andBergemann et al. (2012) on Sr I and II, where large NLTE cor-rections were found for Sr I, while the largest impact on Sr IIstems from the NLTE corrections to the stellar parameters.Article number, page 7 of 18 &A proofs: manuscript no. MoRucorfinallang
Table 4.
Stellar abundances of Mo and Ru for giant stars.
Giants T log g ξ [Fe/H] [Mo / Fe] [Ru / Fe] [K] [km/s]BD-01 2916 4480 a a − .
99 0 .
45 0 . BD+8 2856 4600 a a − .
09 0 . cp . + BD+30 2611 4238 0.50 a − . − .
05 0 . BD+42 621 4725 a a − .
48 0 .
20 0 . BD+54 1323 5213 2.01 c − .
64 0 .
15 0 . CS22890-024 5400 2.65 a − . < . ... CS29512-073 5000 a a − . ... < . CS30312-100 5200 2.35 a − . ... ... CS30312-059 5021 1.90 a − . ... ... CS31082-001 b a − .
81 0 .
94 1 . HD74462 4590 1.84 c − .
48 0 .
50 0 . HD83212 4530 1.21 c − .
25 0 . + . + HD88609 b c − .
87 0 .
34 0 . HD108317 5360 2.76 1.2 − .
11 0 .
20 0 . HD110184 4450 a c − .
40 0 .
25 0 . HD115444 b c − .
00 0 .
36 0 . HD122563 b a a − .
81 0 .
14 0 . HD122956 4700 1.51 1.2 − .
45 0 .
30 0 . HD126238 4900 1.80 1.5 − .
92 0 .
18 0 . HD126587 4700 a c − . < . < . HD128279 5200 a a − . − .
10 0 . HD165195 4200 c c − .
10 0 .
26 0 . HD166161 b a c − .
25 0 .
47 0 . HD175305 5100 2.70 1.2 − .
38 0 . > . ∗ HD186478 4730 1.50 c − .
42 0 .
22 0 . HD204543 4700 0.80 a − .
84 0 .
24 0 . HE0315+0000 5200 2.40 a − . ... < . HE0442-1234 4530 0.65 a − . − .
05 0 . HE1219-0312 4600 1.05 a − . ... ... ∗ Spike in spectral line. + Larger uncertainty in the abundance due to blends. cp Continuum placement uncertain. a , b , c : a) T is derived from excitation potentials, log g from ionisation equilibrium, b) the star has a special r-processpattern, c) uncertain colour, dereddening, or parallax lead to adjustment according to a).Table 5 shows Mo abundances derived from the blue3864 Å and the red 5506 Å line. Generally these two linesyield exactly the same abundance (or values within 0.02 dexor 0.14 dex when an upper limit is derived). This goodagreement indicates that the abundances derived from theblue Mo line have not been biased by, for instance, contin-uum placement or line blending. We choose to list the blueand red Mo abundance in different columns to allow a directcomparison between our blue line values and other metal-poor studies, as well as future studies of the red line carriedout as part of surveys such as GES. Based on the detectionsof Mo from the red 5506 Å line, there seems to be a cut instellar parameters for which Mo can be detected. These in-dicate that Mo can in general be detected in cool and warmdwarfs down to [Fe/H] ∼ − . , upper limits down to − dex,and if the star is enhanced or cool ( < K), the line isuseful down to [Fe/H] ∼ − . . The values presented in ourfigures are the abundances from the blue 3864 Å Mo I line. Table 6.
Formation processes of Mo and Ru isotopes (Dauphaset al. 2004). ‘P’ indicates that the isotope is created only by ap-process, similarly for ‘s’ and ‘r’, while ‘r+s’ means that theisotope can be created by both r- and s-processes.
Element IsotopeMo 92 94 95 96 97 98 100Ru 96 98 99 100 101 102 104Process p p r + s s r + s r + s r
To trace the nucleosynthetic origin of Mo and Ru we com-pare these elements to others with known formation pro-cesses. It has been suggested that a weak r-process createselements in the range < Z < , and Hansen et al. (2012)showed that Ag, and to some extent Pd, were created bythis second weak r-process. These observations show notonly that there was one universal r-process creating all r-process dominated elements heavier than Zn, but also that aweak r-process interfered with this picture of the trans-iron Article number, page 8 of 18ansen, C. J. et al.: Molybdenum and ruthenium
Table 5.
Stellar abundances of Mo and Ru for dwarf stars.
Dwarfs T log g ξ [Fe/H] [Mo / Fe] [Mo / Fe] [Ru / Fe] [K] [km/s] (3864A) (5506A)
BD+09 2190 6450 a − . ... ... ... BD-13 3442 6450 4.20 a − . ... ... ... CD-30 18140 6340 4.13 1.0 − . ... ... < . CD-33 3337 5952 3.95 1.4 − . ... ... . CD-45 3283 5657 c − .
99 0 . < .
25 0 . CD-57 1633 5907 4.26 1.1 − . ... ... . + HD3567 6035 4.08 1.5 − .
33 0 . ... . HD19445 5982 4.38 1.4 − . ... ... . HD22879 5792 4.29 1.2 − . ... ... . HD25704 5700 4.18 1.0 − .
12 0 . ... . HD63077 5629 4.15 0.9 − .
05 0 . ... . HD63598 5680 4.16 0.9 − . ... < . . HD76932 5905 4.08 1.3 − .
97 0 . ... . HD103723 6128 4.28 1.5 − .
85 0 . ... . HD105004 5900 a c − . < . ... . HD106038 b − .
48 0 . ... . HD111980 b − .
31 0 . < . . HD113679 5759 4.04 0.9 − .
63 0 .
22 0 . . HD116064 5999 4.33 1.5 − . ... ... < . HD120559 5411 4.75 0.7 − .
33 0 . ... . HD121004 5711 4.46 0.7 − . ... . . HD122196 6048 3.89 1.2 − . ... ... . HD126681 b − .
28 0 .
62 0 . . HD132475 5838 3.90 1.5 − .
52 0 . ... . HD140283 5738 3.73 1.3 − . ... ... ... HD160617 6028 3.79 1.3 − . ... ... . HD166913 6155 4.07 1.5 − .
30 0 . ... . HD175179 5758 4.16 0.9 − .
72 0 . cp < . . HD188510 5536 4.63 1.0 − .
58 0 . ... . HD189558 5712 3.79 1.2 − .
18 0 . < .
68 0 . HD195633 6005 3.86 1.4 − . ... ... . HD205650 5842 4.49 0.9 − .
19 0 . ... . HD213657 6208 3.78 1.2 − . ... ... ... HD298986 6144 4.18 1.4 − .
48 0 . ... . G005-040 5766 4.23 a − .
93 0 . < .
35 0 . G013-009 6416 3.95 1.4 − . ... ... ... G020-024 6482 4.47 1.5 − . ... ... < . G064-012 6459 4.31 c − . ... ... ... G064-037 6494 3.82 c − . ... ... ... G088-032 6327 3.65 1.5 − . ... ... ... G088-040 5929 4.14 1.4 − .
90 0 . ... . G183-011 6309 3.97 1.0 − . ... ... ... + Larger uncertainty in the abundance due to blends. cp Continuum placement uncertain. a , b , c : a) T is derived from excitation potentials, log g from ionisation equilibrium, b) the star has a special r-processpattern, c) uncertain colour, dereddening, or parallax lead to adjustment according to a).elements. Here we explore if this process also contributes tothe formation of Mo and Ru.We start by comparing the stellar derived elementalabundances of Mo and Ru to abundances of Sr, Zr, Pd,Ag, Ba, and Eu derived by Hansen et al. (2012). The abun-dances of these six elements have been derived using thesame method and codes for the analysis as the ones appliedhere. Thus, it is a very homogeneous analysis, which al-lows a direct comparison between the previously publishedabundances and the ones derived here. To ease the comparison made in Figs. 6 to 10, the for-mation process of each element is listed in Table 7. Thevalues listed are from Arlandini et al. (1999), however, inBisterzo et al. (2014) most of the s-process elements haveincreased by ∼ % except for Zr, for which the s-processfraction has decreased to ∼ %. Article number, page 9 of 18 &A proofs: manuscript no. MoRucorfinallang
Fig. 4.
Top: [Mo/Fe] as a function of metallicity. Dwarf starsare shown as black filled circles, giants as filled red triangle,while comparison stars from Peterson (2013) are shown as greenpluses. Bottom: [Ru/Fe] as function of [Fe/H] using the samesymbols as just described. ‘Others’ is a sample comprised datafrom: Sneden et al. (2003); Ivans et al. (2006); Allen & Portode Mello (2007); Honda et al. (2007); Roederer et al. (2010a,b);Peterson (2011); Roederer et al. (2012); Johnson et al. (2013);Roederer et al. (2014).
Table 7.
Formation process of Sr – Eu with percentages fromArlandini et al. (1999).
Element Process (in %) CommentSr 85% s Mostly weak sZr 83% s Mixed: r+s+weak r & sMo 50% s Mixed: p+r+sRu 32% s Mostly weak rPd 46% r Partial weak r contributionAg 79% r Predominantly weak rBa 81% main s Main r at low [Fe/H]Eu 94% main r Always main r
Table 8.
Properties, number of members, and median metallic-ity of automatically optimised clusters of data.
Elements ‘metal-poor cluster’ ‘metal-rich cluster’members [Fe/H] members [Fe/H]Mo – Sr 15 − . − . Mo – Zr 16 − . − . Mo – Ru 15 − . − . Mo – Pd 14 − . − . Mo – Ag 15 − . − . Mo – Ba 7 − . − . Mo – Eu 13 − . − . Ru – Sr 22 − . − . Ru – Zr 23 − . − . Ru – Pd 16 − . − . Ru – Ag 18 − . − . Ru – Ba 15 − . − . Ru – Eu 17 − . − . Interpreting the linear trends
To extract the similarity in formation processes, absolute(log ǫ ) abundances of Mo and Ru are compared to othertrace elements. If the two compared elements are formedby the same process, we expect to find a 1:1 correlation;that is, the fitted line should have a slope of 1.0. One indi-cation of several, competing formation processes is a largerstar-to-star scatter. This will be expressed as a larger un-certainty on the fitted slopes. The trends (lines) have beenfitted using a linear least-squares method that, when tested,turned out to be very robust. The least-squares method wastested against a robust least absolute deviation method anda minimum χ fit. All yielded the same results within theuncertainty of the fit. However, when removing a few starsfrom the sample, the linear least-squares method turnedout to be more consistent in yielding robust slopes. Up-per limits have been removed to ensure a cleaner trend ofthe two elements compared. Since the comparison has beencarried out on absolute abundances, no obscuration fromother elements such as Fe or the choice of solar abundancehave been introduced. Owing to the small abundance un-certainties ( ∼ . dex), we can require very tight correla-tions between the two elements. Furthermore, to accept thecorrelation, we also require that the star-to-star scatter islow. This is quantified and constrained through a 1 σ un-certainty of the fitted lines being similar to or less than theabundance uncertainty of 0.15 dex. However, if the gas hasbeen mixed or diluted with contributions from other forma-tion processes, the correlation is weaker , and the slope willdeviate from 1.0 by ∼ . or more. Phrased differently, ifthe slope is below 0.85 or above 1.15, one of the two ele-ments must be enriched by a different secondary process,which will break the 1:1 correlation.Two different fitting approaches have been followed: 1)We fit dwarfs and giants separately, since we expect themto be affected differently by NLTE effects, and 2) we carryout an automated cluster analysis, where different weightsare assigned to the data, which ensures that the data withsimilar properties are placed in a cluster.In the first case, the offsets in the slopes fitted to thedwarfs and giants can also be taken as an expression ofhow important NLTE effects might be for the two elements Article number, page 10 of 18ansen, C. J. et al.: Molybdenum and ruthenium
Fig. 5.
Single trends fitted to Mo abundances vs Sr - Eu (left), and Ru vs Sr - Eu (right). Three asterisks indicate a 1:1 correlation. shown. In the second case, the properties of the data havebeen taken into account. For a given number of clusters,the centre is determined, and the data is placed in a clus-ter via a minimisation of the data point’s distance to thecentre. The clustering was done in IDL using the routines‘cluster_wts’ and ‘cluster’. These routines uses a k-meansclustering where the initial clusters are chosen randomly,and data points are moved between clusters by minimisingthe variability within the cluster, thereby increasing the dif-ference between clusters. Since random clusters are initiallyassigned to the data every time the routine runs, differ-ent results could be obtained, especially for scattered datapoints. However, this is not the case, and in the ∼ testruns, the same clusters have always been obtained for thissample.We find that at most two clusters are needed to describethe data, otherwise the clusters get very small and are lesssignificant with respect to a physical formation process. Theabundances can be well fitted using one or two clusters.Figure 5 shows a summary of the correlations between Moand Sr to Eu and between Ru and Sr to Eu. From this figure,Mo is seen to closely correlate with most of the elements,except for Ag. Thus, Mo seems to have a very mixed origineven at low metallicities. Ruthenium, on the other hand,shows the tightest correlation with Ag, Pd, and Zr; it hasless in common with the production channel of Eu, andhardly anything to do with the s-processes creating Sr andBa. In most cases, the lines fitted are associated with verylow uncertainties ( < . ), except for the trends betweenMo and Sr or Ag and Ru and Sr. The weaker trend betweenMo and Ag plus the large uncertainty associated with thelinear fit between weak r-process elements and Mo couldindicate that the weak r-process might affect Sr, Mo, Ru,and Ag differently at different times or metallicities.We therefore explored what happens when two clustersare fitted instead of just one. For each of the two clusters, we calculated the median of the metallicity in the cluster.The cluster size and median metallicity are listed in Table 8.In the cluster fitting case we generally find significantly dif-ferent [Fe/H] median values. In most cases, the metal-poorcluster consists of metal-poor giants and dwarfs and one ortwo more metal-rich dwarf stars. The ‘metal-rich’ (median[Fe/H] > − . ) cluster mainly consists of more metal-richdwarf stars with a few giants or a few metal-poor stars (2– 5 stars with [Fe/H] < − depending on the pair of ele-ments). Assigning the data points to clusters seem to leadto a natural division of a more metal-poor sample versusa metal-richer sample, where the former most likely willbe dominated by the r-process and the latter contaminatedmore by the s-process. Thus, by splitting the sample intosub-clusters we seem to gain more information on the un-derlying physical formation processes, rather than by look-ing at dwarfs and giants separately. The clusters providea stronger lead on the behaviour of the formation process,and we proceed by fitting linear trends to the clusters untilfurther notice on the actual NLTE abundance behaviour.The trends fitted to the giants generally agree with thoseof the metal-poor cluster. This makes sense, since the metal-poor cluster mainly contains metal-poor giant stars. Trends from figures
All the fitted lines, slopes, and intersections with the y-axisare given in each of the panels in Figs. 6 to 10. Immedi-ately below these linear equations, the 1 σ uncertainty onthe slope and intersection is given in parenthesis.We start by comparing Mo and Ru in Fig. 6, where wefind correlations within the uncertainty at both higher andlower metallicity. This is not surprising since the isotopes ofthese elements are created by similar processes. However,to pin down the reason for these trends and understand theinfluence of the weak r-process, we need to compare Mo and Article number, page 11 of 18 &A proofs: manuscript no. MoRucorfinallang
Fig. 6.
Top: Weak correlation between Mo and Ru. Dwarfs(filled black circles) and giants (filled red triangles) with fittedlines plotted on top. Bottom: Linear trends fitted to Mo and Ruabundances in the automatically assigned clusters. Red pointsare generally more metal-poor than black points.
Ru to elements with well-determined formation processes.In the following Figs. 7 to 10 the red triangles are generallythe more metal-poor stars, while the black circles are moremetal-rich.From panels a) and b) in Fig. 7 we find almost 1:1 corre-lations between Mo and Sr and Mo and Zr, respectively, atlower metallicity. This could indicate that weak s-processyields have been incorporated in stars with metallicitiesbelow [Fe/H] = − . . The trends are clean (a slope of0.94 ± . ) and the uncertainties low. On the other hand,from Table 7 we see that 15% of Sr is created by a pro-cess that is different from the weak s-process. It is differentfrom the weak r-process (see lower panels of Fig. 7), but itcould be a sort of lighter element primary process (LEPP),such as an α -process or a ν p-process (Fröhlich et al. 2006).Another formation channel is the charged-particle processdescribed in Qian & Wasserburg (2008). At low [Fe/H] ei-ther of these processes could be driving the enrichment ofelements from Sr up to Mo instead of the weak s-process. Athigher metallicities the slopes in the two top panels clearlydeviate from unity, and the uncertainty (star-to-star scat-ter) is large. This could indicate that there are several for- Fig. 7.
Absolute abundance of Mo vs Sr, Zr, Pd, and Ag.Article number, page 12 of 18ansen, C. J. et al.: Molybdenum and ruthenium mation processes creating Mo at higher [Fe/H]. One optionwould be the p-process or the earlier mentioned α -/ ν p-process, which would explain the correlation between Moand Ru at higher [Fe/H] since their lightest isotopes arecreated by a p-process. In the following two panels (Fig. 7c) and d), Mo is seen to correlate with Pd at higher metal-licity, indicating that both elements may receive a largecontribution possibly from a weak or main s-process. How-ever, the remaining trends in these panels do not show a1:1 correlation, but are instead described by large star-to-star abundance scatter resulting in uncertain linear trends.Based on this, we do not believe that Mo is predominantlycreated by a weak r-process, but it may receive a minorcontribution.Continuing to compare Mo to Ba in Fig. 9 a), we seethat the two elements show a tight correlation at higher[Fe/H], but it clearly deviates from a line with slope 1.0.This indicates that a different process like the weak s- or thep-process interferes and creates Mo in addition to the mains-process. At low [Fe/H], the cluster is small and the datais scattered. This could also indicate that this study onlyincludes a few main s-process dominated stars below [Fe/H]= − or that another process is interfering. A larger sam-ple of both metal-poor and metal-rich stars containing Moand Ba would be needed to clarify this. The largest amountof Mo is also not created by the main r-process, which isresponsible for the production of Eu (see Fig. 9 panel b).Molybdenum and Eu show a large scatter and uncertaintyand a deviation from a 1:1 correlation when split into sub-clusters. This could, as mentioned above, indicate that evenlarger samples are needed. A single cluster indicates thatMo and Eu correlate almost 1:1 with a low scatter. How-ever, differences in formation processes or traces of a possi-ble late onset of a process (e.g. the main s-process) seem tobe erased when a single cluster is enforced. This makes ithard to tell if small number statistics are playing a domi-nant role, or if different processes are setting in at differenttimes. Increasing the sample of stars with Eu, Ba, and Moin both metal-poor and metal-rich stars would help clarifythis issue.Ruthenium is different from most weak s-process-dominated elements at higher metallicity (see Fig. 8 panela), b), and less so in c); cf. Table 7). This is expressed inslopes that differ by more than 0.25 from unity and un-certainties of up to 0.41. This uncertainty is the largestderived and is found between Ru and Sr in both one andtwo clusters. This picture changes at lower metallicity, andif the weak s-process is already efficient at this low metallic-ity, these trends and differences between higher and lowermetallicity become hard to explain. Again, we note thata LEPP, α - or ν p-process may be responsible for the for-mation of Sr - Ru, in which case the issue with the weaks-process is circumvented.Very direct and clean almost 1:1 trends are found be-tween Ru and the weak r-process element Ag (Fig. 8 d).This shows that Ru receives a dominant contribution fromthe weak r-process regardless of metallicity, and the influ-ence of the weak r-process may also play a role in the cor-relations we find at lower metallicity between Ru and Zr,but less so between Ru and Sr. This could indicate that Zris not mainly produced by the s-process, as predicted byArlandini et al. (1999), but rather ∼ / r/s, as recentlyindicated in Bisterzo et al. (2014). Fig. 8.
Absolute abundance of Ru vs Sr, Zr, Pd, and Ag.Article number, page 13 of 18 &A proofs: manuscript no. MoRucorfinallang
Fig. 9.
Mo abundances compared to those of Ba and Eu.
Finally, Fig. 10 a) shows Ru versus Ba, where the abun-dances of these elements do not correlate at any metallic-ity, and the slopes deviate the most from 1.0, so the mains-process has the lowest influence on the Ru abundances.This is seen in one and two clusters as well as in dwarfs andgiants. Panel b) of the same figure shows a large star-to-starscatter and slopes different from unity. Both indicate thatthe main r-process is not dominating the Ru production,though it is still contributing more than the main s-processis. Even when considering that Ba could be produced by amain r-process at low [Fe/H] (cf. Table 7), we still find cleardifferences between Mo and both heavy elements (Ba andEu), as well as between Ru, Ba, and Eu. This shows thatMo and Ru are not predominantly produced by the main r-process and that the process driving their correlations maybe a sort of LEPP. This process is clearly different from themain r-process, which is responsible for the formation andslopes of Eu, and also for Ba at low [Fe/H].To ease comparison of the cluster slopes, they have beensummarised in Fig. 11. These have been shifted for the sakeof clarity to go through the point (1,1), but all the slopeshave been preserved. This also allows a very direct compar-ison to the slopes obtained from only one cluster with oneline fitted to the full sample (see Fig. 5).With the adopted log gf values, the focus on continuumplacement and blends, the elemental abundances are as pre-cise as currently possible, and we consider the trends trust-
Fig. 10.
Ru abundances compared to those of Ba and Eu. worthy. We are aware that larger sample sizes will mostlikely improve the trends and make the according resultsmore robust.Generally, at high [Fe/H], Mo is seen to correlate withRu and Pd, which in the case of Ru could mean a consid-erable p-process contribution . At lower [Fe/H] Mo cor-relates directly with Sr and Zr. This could point towardsearly weak s-process contributions in which case the weaks-process range goes up to and includes Mo but not Pd;however, this may also be true for the α - or ν p-process,for instance. If the weak s-process is the source, it wouldstrengthen the scenario described by Pignatari et al. (2013).In one case, Pignatari et al. (2013) show that Mo and Ruwould require a very high s-process rate leading to a dis-agreement between weak s- and main s-process, while the p-isotopes of Mo and Ru can be reproduced via a ‘cs-process’in massive stars (M ∼ M ⊙ ) or via a ν p-process (Fröh-lich et al. 2006). Ruthenium is seen to be a less convolvedelement than Mo, and Ru is mainly formed by the weakr-process as is Ag.The study by Lorusso et al. (2011) combining experi-ments and theory shows that the formation process of Ruis unlikely to be an rp-process. The exact environment andsites of these processes are far from resolved, and we leavefurther discussions to a future paper (see e.g. Fröhlich etal, 2014 in prep). The p-process may be metallicity sensitive and show a sec-ondary nature according to Pignatari et al. (2013)Article number, page 14 of 18ansen, C. J. et al.: Molybdenum and ruthenium
Fig. 11.
Comparison of slopes fitted to the elements given in the legend. Top panels show Mo with respect to other key elements(Sr–Eu), while the bottom panels show Ru. Median metallicity of the cluster is stated in each figure.
5. Discussion
Even with elemental abundances we are capable of distin-guishing between the formation processes and find the onethat donate the dominant amount to each of the elements.However, considering isotopic abundances of presolar grainsallows a much more detailed and direct comparison betweenthe isotopic abundance and associated formation process.
Presolar grains
The presolar grain abundances are taken from the preso-lar database (Hynes & Gyngard 2009) at St. Louis Univer-sity . We compare abundances of the SiC grains withinthe database that have known abundances of Mo and min-imum one of the other heavy elements shown in Sect. 4.1.Presolar silicon carbide grains are the best studied presolarmineral phase. This is due to the relatively high abundancein primitive meteorites and to the size of the SiC grains,which allows isotopic analyses, of the major and of manytrace elements, of individual grains. The presolar SiC grainsare categorised based on the abundant elements C and Si,as well as on the trace elements that are present within theminerals. Isotopic data exist for N, Mg, Ca, Ti, the noblegases, and heavy refractory elements (e.g. Zr, Mo, La, Nd,Sm, Ru, and Ba). Based on the isotopic composition of C,N, Si, and the abundance of radiogenic Mg, six differentpopulations of SiC grains are discerned: the mainstreamgrains, which make up the majority of the grains (around90% of the total), and the minor types A, B, X, Y, and Z.For our study, the X grains are of particular interest becausetheir C- and Si-isotopic compositions can be explained by http://presolar.wustl.edu/pdg and E. Zinner priv. comm. mixing of matter from the C- and Si-rich zones in a type IISN. The X-grains are characterised by enrichments (rela-tive to solar isotopic abundance) in C (most grains), N,and Si. Many of the grains have isotopic over-abundancein Mg (Hoppe et al. 1994; Virag et al. 1992; Zinner et al.1991), Ca (Nittler et al. 1996), and Ti (Hoppe et al.2000; Amari et al. 2000), which likely came from the ra-dioactive decay of Al (t / ∼ . Myr), Ti (t / ∼ yr), and V (t / ∼ d) after grain formation. We referto Lodders & Amari (2005), and Hoppe et al. (2010) forfurther details on presolar grains.The abundances of presolar grains are given as ratios onthe δ -abundance scale between the meteoritic grain and astandard, δ i X = (cid:18) ( i X/ j X ) grain ( i X/ j X ) standard − (cid:19) · , (1)where X is the element, and i and j different isotopes. Thestandard is often taken from a terrestrial sample or thereference values from Anders & Grevesse (1989). The totalabsolute meteoritic abundance cannot be extracted fromthese measurements, so we are forced to use relative isotopicabundances, which among other things can describe thefraction of r-/s-process present in the grain. Assuming thenatural ratios of p-, r-, and s-isotopes are universal, thisratio should be descriptive of both the grain and the totalmeteorite, as well as the stars. Furthermore, neither Zr norBa are siderophile elements (Mo only slightly so), nor arethey strongly separated between the gas and solid phases.Recently, Mann et al. (2012) have shown that Ru wasnot partitioned, not even under high pressure or temper-ature. Both Mo and Zr are refractory elements and eventhough Ba is less refractory, it is still far from being volatile. Article number, page 15 of 18 &A proofs: manuscript no. MoRucorfinallang
It therefore seems most natural that their abundances arerepresentative of the original gas composition. Actually,Lodders (2010) investigated how the siderophile nature af-fected log ǫ Mo abundances in the Sun, now and just af-ter the formation of the Sun. The difference can be foundby comparing the solar Mo in their Tables 3 and 6, whichshows that the difference is 0.07 dex and no difference forRu. However, here we have chosen not to correct for thesechanges in Mo or other elements, since we have a mixtureof dwarfs and giants, and they span a broad range of ages.The giants have a convective atmosphere, and we do notknow how this would affect these estimates. Furthermore,we also have stars that are older than the Sun. Combiningthe variation in ages and convection, this would lead to adifference in the solar reference value of Mo. In either casethis value will most likely remain small and stay below theadopted uncertainty. Therefore, it will not have any impacton the correlation trends shown here, and we find it safe todiscard this.We now focus on Zr, Mo, and Ba because Zr and Baseem to be some of the most studied elements in combina-tion with Mo. The most abundant isotope is used as the ref-erence in the presolar database, and the ‘j’ isotope in Eqs.1 and 2, is Zr (s-dominated), Mo (s-only), and
Ba(s-only) when we take the data from the presolar database.This means that we have a pure s (or at least s dominated)denominator in the relative log ǫ -ratio in Eq. (2).To obtain the relative ratios, we first calculate the puregrain ratio of i X j X grain by removing the Earth reference fromthe δ -ratio in Eq. 1. The numerator isotopes ‘i’ for Zr are 91,the most s-process-dominated Zr isotope, and 96 (r-only).Similarly we select r- and s-only isotopes for Mo and Ba tocalculate r/s and s/s ratios. For Mo the two numerator iso-topes are 98 (second most s-dominated isotope after 96) and100 (r-only), while for Ba isotope 135 (most r-dominated Baisotope existing) and 138 (second most s-dominated isotopeafter 136) were chosen. The choice of isotopes was made inaccordance with Table 1 in Sneden et al. (2008). The log ǫ -ratios are given by Eq. 2 and plotted on top of the stellarabundances in Fig. 12: log ǫ = log i X j X grain i X j X A & G + 1 . . (2)Here we selected the meteoritic abundances (A&G) fromAnders & Grevesse (1989) to ensure a consistent referencescale. The constant (1.554) is a conversion factor taken fromAnders & Grevesse (1989), which converts the abundancesfrom a meteoritic Si atom scale to an astrophysical H atom scale. The δ -abundances of SiC are originally fromNicolussi et al. (1997) and the SiC X from Pellin et al.(2006).In the top panel of Fig. 12 we have shown the total stel-lar abundance of Mo, Zr, as well as the average of the frac-tions of r- and s-processed material derived from presolargrains. The r/s and s/s ratios are seen to agree well with thetotal mixed stellar abundance derived from dwarfs with ametallicity of [Fe/H] = − . to − . . We have also includedthe Sun (abundances from Anders & Grevesse 1989) in Fig.12 to guide the following discussion.The presolar SiC grain r/s ratios match the abundanceof the metal-poor dwarfs down to [Fe/H] ∼ − . . Above Fig. 12.
Comparing r/s and s/s ratios in presolar grains to starsfrom our samples including the Sun as reference. [Fe/H] = − . , the r/s ratios must therefore be the same ev-erywhere in the Galactic ISM, and the gas composition thusseems homogeneous. The smooth decreasing trend down to0.0 in log ǫ (Mo) could be taken as an expression of this,while a deviation from the trend is seen below 0.0 (Fig.12 bottom panel). It could indicate that heterogeneity ismainly found below log ǫ (Mo) ∼ , which for this sam-ple corresponds to [Fe/H ] < − . . Therefore we cannotsupport a heterogeneous nebula at these high metallicities ∼ − . to − . (as in Dauphas et al. 2004 and Burkhardtet al. 2011). The suggested decoupling between p- and r-processes (or just between r-processes; Chen et al. 2004) ismore likely based on our stellar abundance correlations inSect. 4.1.The SiC X grains (open square) are seen to have aslightly larger r/s fraction than the mainstream SiC (opendiamond see Fig. 12). This agrees with SiC X grains beingenriched by supernovae type II (Hoppe et al. 1996). Thatthe s-dominated Mo ratios are greater than the correspond-ing r/s ratios in mainstream SiC grains shows that these SiCgrains are depleted in r-process, which agrees with Pellinet al. (2006) and Nicolussi et al. (1997). In the mainstreamSiC grains, the isotopic s-abundances are larger than thestellar abundances. This also agrees with the facts that SiCgrains are generally thought to be directly enriched by anAGB stars and that the s-process created isotopes mightbe formed at the expense of the r-isotopes. Since the grainsare more s-process-enriched than the stars, it could indicate Article number, page 16 of 18ansen, C. J. et al.: Molybdenum and ruthenium that the gases in stars are more diluted or differently mixedthan the SiC grains. The mixing processes need to be bet-ter constrained, both within the AGB stars, which directlyaffects the SiC grains, and in the ISM where yields from SNand AGB will be incorporated into stars. Until the mixingis better understood, stronger conclusions cannot be drawnbased on either elemental or isotopic abundances. The me-teorites seem to have formed from a gas similar to that ofthe stars at [Fe/H] = − . to − . . This agrees with thepresolar grains being made prior to (at a lower metallicitythan) the Sun.From the bottom panel of Fig. 12, both SiC grains areseen to contain more s-process- than r-process-created ma-terial. This is in good agreement with what we found in thetop panel. Furthermore, the abundance compositions followa Galactic chemical evolution scheme, where the mixtureof supernova r- and AGB star s-process material enrich thedwarfs and giants around and above − . in [Fe/H], and yetstay below the solar ratios. In general, the presolar grainsare slightly more enhanced in s-process material than themetal-poor dwarfs around [Fe/H] = − . to − . . DespiteBa being a less refractory element than Mo, and Mo moresiderophile than Ba, the Ba/Mo ratio from the grains seemsto match that of the stars . This could help put constraintson how efficient the mixing processes are in mixing AGBand/or SN gases into later generations of stars.This comparison of stars and grains show that bothseem to share the same mixture of r/s-gases and there-fore also formation processes around [Fe/H] ∼ − . andabove. Figure 12 indicates that the stars around this metal-licity show a mixture of both r- and s-processes, whichmeans that s-process traces are seen down to (and likelybelow) [Fe/H] = − . . Below this metallicity Roederer et al.(2010a) found little or no s-process in their sample stars,and Peterson (2011) conclude that s-process yields ‘werelargely absent’ in their sample. This is in concordance withTravaglio et al. (2004) where the weak s-process contribu-tion to halo stars is found to negligible. Our comparison tograins indicate that this cut in metallicity may be moveddown in metallicity to somewhere between [Fe/H] = − . and − . . The lower value we base on the increasing abun-dance star-to-star scatter found in Hansen et al. 2012 atthis metallicity. Furthermore, this value is in good agree-ment with the average metallicity ( − . ) we find from thestars below log ǫ (Mo) = 0 in Fig. 12.
6. Summary and conclusion
We summarise the outcome from the stellar abundancesfirst, and then conclude on both stellar and meteoritic abun-dances. Ruthenium is in all cases seen to correlate almostperfectly with silver, and this provides a strong observa-tional indication of ruthenium being created by the weakr-process. The weak r-process is dominating the formationof Ru. Smaller amounts of Ru are created by the p- and r-processes, as well as the weak s-process. The main s-processis in this connection the poorest donor.From Fig. 11, Mo is seen to have more in common withthe lighter elements than the heavy elements. Based onthese observations, Mo can be considered as a highly mixed This could indicate that the corrections from Lodders (2010)are indeed small or cancel out. element, where contributions from p-, main, and weak s-processes are all mixed with smaller contributions from themain r-process. The influence of the weak r-process is of lessimportance. The weak s-process seems to be able to formMo. This could confirm the extension of the weak s-processto include Mo, as found by Pignatari et al. (2010, 2013), de-spite the mentioned model complications. The exact onsetof the weak s-process is hard to trace.According to Travaglio et al. (2004) the weak s-processcontributes little to the metal-poor halo stars owing to themetallicity dependence of the process, and they point to-wards a LEPP origin for Sr–Zr instead of the s-processes.At low [Fe/H] the lighter elements studied here may verywell be produced by a ν p-, α -, charged-particle, or someother primary process rather than the weak s-process. Ineither case, this process should be different from the weakr-process creating Ru – Ag.To understand the formation of Mo and Ru in greaterdetail at low metallicity, the metallicity dependence of p-,s-, cs-processes, etc. need to be understood first. Moreover,the mass of the production site is another important quan-tity to constrain, in addition to the metallicity, and thecombined behaviour of yields need to be explored in Galac-tic chemical evolution models.By studying the two clusters instead of one, the abun-dance scatter and uncertainty in the fitted line between Moand Sr is found at higher metallicity, while between Moand Ag the scatter is found at lower metallicity. This couldindicate that different processes are dominating the Mo pro-duction as a function of time or metallicity (cf. Sect. 4). Topin down the subtleties of the formation processes at thislevel, isotopic abundances, or even larger samples spanningbroad Fe, Ru, and Mo abundance ranges are needed.The r-/s-process mixture in presolar grains agree wellwith the chemical composition of the dwarf stars around[Fe/H] = − . to − . . This could indicate that both grainsand stars around and above [Fe/H] = − . are mixed well(as also seen from Galactic chemical evolution trends),and we can therefore not support a heterogeneous preso-lar nebula. An inhomogeneous ISM is only expected atlower metallicities. A possibility for abundance anomaliesor differing r/s or s/s ratios therefore ought to arise fromdifferences in nucleosynthetic origin. Another possibilitythat might explain anomalies could be related to mea-suring techniques, or problems with e.g. fractionation. Amain s-process yield of Mo and Ba is seen in stars with[Fe/H] = − . . The difference between the total stellarabundance and the s-only isotopes from the grains could in-dicate that AGB yields are less efficiently mixed into starsthan presolar grains. However, we need a better mixing de-scription and hyperfine splitting for Mo to derive isotopicstellar Mo abundances to accurately probe this. Acknowledgements.
This work was supported by Sonderforschungs-bereich SFB 881 "The Milky Way System" (subproject A5) of theGerman Research Foundation (DFG). The Dark Cosmology Centreis funded by the Danish National Research Foundation. We wouldlike to thank the anonymous referee for constructive comments. CJHthanks U. G. Jørgensen, H.-P. Gail, and C. Fröhlich for discussion.CJH also thanks L. Nittler, and E. Zinner for guidance and access tothe presolar database. This research has made use of NASA’s Astro-physics Data System, the SIMBAD database, operated at the CDS,Strasbourg, France, and the Two Micron All Sky Survey, which isa joint project of the University of Massachusetts and the InfraredProcessing and Analysis Center/California Institute of Technology,funded by the National Aeronautics and Space Administration andthe National Science Foundation.
Article number, page 17 of 18 &A proofs: manuscript no. MoRucorfinallang
References
Alexander, C. M. O., Boss, A. P., Keller, L. P., Nuth, J. A., & Wein-berger, A. 2007, Protostars and Planets V, 801Allen, D. M. & Porto de Mello, G. F. 2007, A&A, 474, 221Amari, S., Anders, A., Virag, A., & Zinner, E. 1990, Nature, 345, 238Amari, S., Zinner, E., & Lewis, R. S. 2000, Meteoritics and PlanetaryScience, 35, 997Anders, E. & Grevesse, N. 1989, Geochim. Cosmochim. Acta, 53, 197Andrievsky, S. M., Spite, F., Korotin, S. A., et al. 2011, A&A, 530,A105Arcones, A. & Montes, F. 2011, ApJ, 731, 5Arlandini, C., Käppeler, F., Wisshak, K., et al. 1999, ApJ, 525, 886Becker, H. & Walker, R. J. 2003, Nature, 425, 152Bergemann, M., Hansen, C. J., Bautista, M., & Ruchti, G. 2012, A&A,546, A90Bernatowicz, T., Fraundorf, G., Ming, T., et al. 1987, Nature, 330,728Bernatowicz, T. J., Amari, S., Zinner, E. K., & Lewis, R. S. 1991,ApJ, 373, L73Bisterzo, S., Gallino, R., Straniero, O., Cristallo, S., & Käppeler, F.2011, MNRAS, 418, 284Bisterzo, S., Travaglio, C., Gallino, R., Wiescher, M., & Käppeler, F.2014, ArXiv e-printsBurkhardt, C., Kleine, T., Oberli, F., et al. 2011, Earth and PlanetaryScience Letters, 312, 390Cescutti, G., Chiappini, C., Hirschi, R., Meynet, G., & Frischknecht,U. 2013, A&A, 553, A51Chen, J. H., Papanastassiou, D. A., Wasserburg, G. J., & Ngo, H. H.2004, in Lunar and Planetary Institute Science Conference Ab-stracts, Vol. 35, Lunar and Planetary Institute Science ConferenceAbstracts, ed. S. Mackwell & E. Stansbery, 1431Croat, T. K., Stadermann, F. J., & Bernatowicz, T. J. 2005, ApJ, 631,976Dauphas, N., Davis, A. M., Marty, B., & Reisberg, L. 2004, Earthand Planetary Science Letters, 226, 465Dekker, H., D’Odorico, S., Kaufer, A., Delabre, B., & Kotzlowski, H.2000, in Proc. SPIE, Vol. 4008, 534Farouqi, K., Kratz, K., Mashonkina, L. I., et al. 2009, ApJ, 694, L49François, P., Depagne, E., Hill, V., et al. 2007, A&A, 476, 935Freiburghaus, C., Rosswog, S., & Thielemann, F.-K. 1999, ApJ, 525,L121Frischknecht, U., Hirschi, R., & Thielemann, F.-K. 2012, A&A, 538,L2Fröhlich, C., Martínez-Pinedo, G., Liebendörfer, M., et al. 2006, Phys-ical Review Letters, 96, 142502Goriely, S., Bauswein, A., & Janka, H.-T. 2011, ApJ, 738, L32Gustafsson, B., Edvardsson, B., Eriksson, K., et al. 2008, A&A, 486,951Halliday, A. N. 2003, Nature, 425, 137Hansen, C. J., Bergemann, M., Cescutti, G., et al. 2013, A&A, 551,A57Hansen, C. J., Primas, F., Hartman, H., et al. 2012, A&A, 545, A31Honda, S., Aoki, W., Ishimaru, Y., & Wanajo, S. 2007, ApJ, 666, 1189Honda, S., Aoki, W., Kajino, T., et al. 2004, ApJ, 607, 474Hoppe, P., Amari, S., Zinner, E., Ireland, T., & Lewis, R. S. 1994,ApJ, 430, 870Hoppe, P., Leitner, J., Gröner, E., et al. 2010, ApJ, 719, 1370Hoppe, P., Strebel, R., Eberhardt, P., Amari, S., & Lewis, R. S. 1996,Science, 272, 1314Hoppe, P., Strebel, R., Eberhardt, P., Amari, S., & Lewis, R. S. 2000,Meteoritics and Planetary Science, 35, 1157Hoppe, P. & Zinner, E. 2000, J. Geophys. Res., 105, 10371Hutcheon, I. D., Huss, G. R., Fahey, A. J., & Wasserburg, G. J. 1994,ApJ, 425, L97Hynes, K. M. & Gyngard, F. 2009, in Lunar and Planetary Insti-tute Science Conference Abstracts, Vol. 40, Lunar and PlanetaryInstitute Science Conference Abstracts, 1198Ivans, I. I., Simmerer, J., Sneden, C., et al. 2006, ApJ, 645, 613Jofre, P., Heiter, U., Blanco-Cuaresma, S., & Soubiran, C. 2013, ArXive-printsJohnson, C. I., McWilliam, A., & Rich, R. M. 2013, ApJ, 775, L27Käppeler, F., Gallino, R., Bisterzo, S., & Aoki, W. 2011, Reviews ofModern Physics, 83, 157Karakas, A. I., García-Hernández, D. A., & Lugaro, M. 2012, ApJ,751, 8Kramida, A., Yu. Ralchenko, Reader, J., & and NIST ASD Team.2013, NIST Atomic Spectra Database (ver. 5.1), [Online]. Avail-able: http://physics.nist.gov/asd [2014, January 15]. NationalInstitute of Standards and Technology, Gaithersburg, MD. Kupka F., Ryabchikova T.A., P. N. S. H. W. W. 2000, Baltic Astron-omy, 9, 590Lai, D. K., Smith, G. H., Bolte, M., et al. 2011, AJ, 141, 62Lee, D.-C. & Halliday, A. N. 2002, Meteoritics and Planetary ScienceSupplement, 37, 85Lewis, R. S., Ming, T., Wacker, J. F., Anders, E., & Steel, E. 1987,Nature, 326, 160Lind, K., Bergemann, M., & Asplund, M. 2012, MNRAS, 427, 50Lodders, K. 2010, in Principles and Perspectives in Cosmochemistry,ed. A. Goswami & B. E. Reddy, 379Lodders, K. & Amari, S. 2005, Chemie der Erde / Geochemistry, 65,93Lorusso, G., Becerril, A., Amthor, A., et al. 2011, Physics Letters B,699, 141Mann, U., Frost, D. J., Rubie, D. C., Becker, H., & Audétat, A. 2012,Geochim. Cosmochim. Acta, 84, 593Mashonkina, L., Christlieb, N., Barklem, P. S., et al. 2010, A&A, 516,A46Masseron, T. 2006, PhD thesis, Observatoire de ParisNicolussi, G. K., Davis, A. M., Pellin, M. J., et al. 1997, in Lunar andPlanetary Institute Science Conference Abstracts, Vol. 28, Lunarand Planetary Institute Science Conference Abstracts, 1021Nittler, L. R. 2003, Earth and Planetary Science Letters, 209, 259Nittler, L. R., Amari, S., Zinner, E., Woosley, S. E., & Lewis, R. S.1996, ApJ, 462, L31Nittler, L. R., Hoppe, P., Alexander, C. M. O., et al. 1995, ApJ, 453,L25Nittler, L. R., O’D Alexander, C. M., Gao, X., Walker, R. M., &Zinner, E. K. 1994, Nature, 370, 443Ott, U. & Hoppe, P. 2007, Highlights of Astronomy, 14, 341Pellin, M. J., Savina, M. R., Calaway, W. F., et al. 2006, in Lunarand Planetary Institute Science Conference Abstracts, Vol. 37, 37thAnnual Lunar and Planetary Science Conference, ed. S. Mackwell& E. Stansbery, 2041Peterson, R. C. 2011, ApJ, 742, 21Peterson, R. C. 2013, ApJ, 768, L13Pignatari, M., Gallino, R., Heil, M., et al. 2010, ApJ, 710, 1557Pignatari, M., Hirschi, R., Wiescher, M., et al. 2013, ApJ, 762, 31Qian, Y.-Z. & Wasserburg, G. J. 2008, ApJ, 687, 272Roederer, I. U., Cowan, J. J., Karakas, A. I., et al. 2010a, ApJ, 724,975Roederer, I. U., Lawler, J. E., Sobeck, J. S., et al. 2012, ApJS, 203,27Roederer, I. U., Marino, A. F., & Sneden, C. 2011, ApJ, 742, 37Roederer, I. U., Preston, G. W., Thompson, I. B., Shectman, S. A.,& Sneden, C. 2014, ApJ, 784, 158Roederer, I. U., Sneden, C., Lawler, J. E., & Cowan, J. J. 2010b, ApJ,714, L123Russell, S. S., Lee, M. R., Arden, J. W., & Pillinger, C. T. 1995,Meteoritics, 30, 399Siqueira Mello, C., Spite, M., Barbuy, B., et al. 2013, A&A, 550, A122Sneden, C., Cowan, J. J., & Gallino, R. 2008, ARA&A, 46, 241Sneden, C., Cowan, J. J., Lawler, J. E., et al. 2003, ApJ, 591, 936Sneden, C. A. 1973, PhD thesis, The University of Texas at AustinStancliffe, R. J., Dearborn, D. S. P., Lattanzio, J. C., Heap, S. A., &Campbell, S. W. 2011, ApJ, 742, 121Travaglio, C., Gallino, R., Arnone, E., et al. 2004, ApJ, 601, 864Travaglio, C., Röpke, F. K., Gallino, R., & Hillebrandt, W. 2011, ApJ,739, 93Virag, A., Wopenka, B., Amari, S., et al. 1992, Geochim. Cos-mochim. Acta, 56, 1715Vogt, S. S., Allen, S. L., Bigelow, B. C., et al. 1994, in Proc. SPIE,Vol. 2198, 362Wanajo, S., Janka, H.-T., & Müller, B. 2011, ApJ, 726, L15Westin, J., Sneden, C., Gustafsson, B., & Cowan, J. J. 2000, ApJ,530, 783Yin, Q., Jacobsen, S. B., & Yamashita, K. 2002, Nature, 415, 881Yong, D., Karakas, A. I., Lambert, D. L., Chieffi, A., & Limongi, M.2008, ApJ, 689, 1031Zinner, E. 1998, Annual Review of Earth and Planetary Sciences, 26,147Zinner, E., Amari, S., Anders, E., & Lewis, R. 1991, Nature, 349, 51Zinner, E. K. 2005, Presolar Grains, ed. A. M. Davis, H. D. Holland,& K. K. Turekian (Elsevier B), 17[2014, January 15]. NationalInstitute of Standards and Technology, Gaithersburg, MD. Kupka F., Ryabchikova T.A., P. N. S. H. W. W. 2000, Baltic Astron-omy, 9, 590Lai, D. K., Smith, G. H., Bolte, M., et al. 2011, AJ, 141, 62Lee, D.-C. & Halliday, A. N. 2002, Meteoritics and Planetary ScienceSupplement, 37, 85Lewis, R. S., Ming, T., Wacker, J. F., Anders, E., & Steel, E. 1987,Nature, 326, 160Lind, K., Bergemann, M., & Asplund, M. 2012, MNRAS, 427, 50Lodders, K. 2010, in Principles and Perspectives in Cosmochemistry,ed. A. Goswami & B. E. Reddy, 379Lodders, K. & Amari, S. 2005, Chemie der Erde / Geochemistry, 65,93Lorusso, G., Becerril, A., Amthor, A., et al. 2011, Physics Letters B,699, 141Mann, U., Frost, D. J., Rubie, D. C., Becker, H., & Audétat, A. 2012,Geochim. Cosmochim. Acta, 84, 593Mashonkina, L., Christlieb, N., Barklem, P. S., et al. 2010, A&A, 516,A46Masseron, T. 2006, PhD thesis, Observatoire de ParisNicolussi, G. K., Davis, A. M., Pellin, M. J., et al. 1997, in Lunar andPlanetary Institute Science Conference Abstracts, Vol. 28, Lunarand Planetary Institute Science Conference Abstracts, 1021Nittler, L. R. 2003, Earth and Planetary Science Letters, 209, 259Nittler, L. R., Amari, S., Zinner, E., Woosley, S. E., & Lewis, R. S.1996, ApJ, 462, L31Nittler, L. R., Hoppe, P., Alexander, C. M. O., et al. 1995, ApJ, 453,L25Nittler, L. R., O’D Alexander, C. M., Gao, X., Walker, R. M., &Zinner, E. K. 1994, Nature, 370, 443Ott, U. & Hoppe, P. 2007, Highlights of Astronomy, 14, 341Pellin, M. J., Savina, M. R., Calaway, W. F., et al. 2006, in Lunarand Planetary Institute Science Conference Abstracts, Vol. 37, 37thAnnual Lunar and Planetary Science Conference, ed. S. Mackwell& E. Stansbery, 2041Peterson, R. C. 2011, ApJ, 742, 21Peterson, R. C. 2013, ApJ, 768, L13Pignatari, M., Gallino, R., Heil, M., et al. 2010, ApJ, 710, 1557Pignatari, M., Hirschi, R., Wiescher, M., et al. 2013, ApJ, 762, 31Qian, Y.-Z. & Wasserburg, G. J. 2008, ApJ, 687, 272Roederer, I. U., Cowan, J. J., Karakas, A. I., et al. 2010a, ApJ, 724,975Roederer, I. U., Lawler, J. E., Sobeck, J. S., et al. 2012, ApJS, 203,27Roederer, I. U., Marino, A. F., & Sneden, C. 2011, ApJ, 742, 37Roederer, I. U., Preston, G. W., Thompson, I. B., Shectman, S. A.,& Sneden, C. 2014, ApJ, 784, 158Roederer, I. U., Sneden, C., Lawler, J. E., & Cowan, J. J. 2010b, ApJ,714, L123Russell, S. S., Lee, M. R., Arden, J. W., & Pillinger, C. T. 1995,Meteoritics, 30, 399Siqueira Mello, C., Spite, M., Barbuy, B., et al. 2013, A&A, 550, A122Sneden, C., Cowan, J. J., & Gallino, R. 2008, ARA&A, 46, 241Sneden, C., Cowan, J. J., Lawler, J. E., et al. 2003, ApJ, 591, 936Sneden, C. A. 1973, PhD thesis, The University of Texas at AustinStancliffe, R. J., Dearborn, D. S. P., Lattanzio, J. C., Heap, S. A., &Campbell, S. W. 2011, ApJ, 742, 121Travaglio, C., Gallino, R., Arnone, E., et al. 2004, ApJ, 601, 864Travaglio, C., Röpke, F. K., Gallino, R., & Hillebrandt, W. 2011, ApJ,739, 93Virag, A., Wopenka, B., Amari, S., et al. 1992, Geochim. Cos-mochim. Acta, 56, 1715Vogt, S. S., Allen, S. L., Bigelow, B. C., et al. 1994, in Proc. SPIE,Vol. 2198, 362Wanajo, S., Janka, H.-T., & Müller, B. 2011, ApJ, 726, L15Westin, J., Sneden, C., Gustafsson, B., & Cowan, J. J. 2000, ApJ,530, 783Yin, Q., Jacobsen, S. B., & Yamashita, K. 2002, Nature, 415, 881Yong, D., Karakas, A. I., Lambert, D. L., Chieffi, A., & Limongi, M.2008, ApJ, 689, 1031Zinner, E. 1998, Annual Review of Earth and Planetary Sciences, 26,147Zinner, E., Amari, S., Anders, E., & Lewis, R. 1991, Nature, 349, 51Zinner, E. K. 2005, Presolar Grains, ed. A. M. Davis, H. D. Holland,& K. K. Turekian (Elsevier B), 17