THE HISTORY OF CHEMICAL ENRICHMENT AND THE SITES OF EARLY NUCLEOSYNTHESIS: CNO ABUNDANCES OF GALACTIC CARBON-ENHANCED METAL-POOR STARS

Transcription

1 THE HISTORY OF CHEMICAL ENRICHMENT AND THE SITES OF EARLY NUCLEOSYNTHESIS: CNO ABUNDANCES OF GALACTIC CARBON-ENHANCED METAL-POOR STARS By Catherine R. Kennedy A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Physics and Astronomy 2011

2 ABSTRACT THE HISTORY OF CHEMICAL ENRICHMENT AND THE SITES OF EARLY NUCLEOSYNTHESIS: CNO ABUNDANCES OF GALACTIC CARBON-ENHANCED METAL-POOR STARS By Catherine R. Kennedy This dissertation focuses on abundance analyses of carbon-enhanced metal-poor (CEMP) Galactic halo stars. Different methods for determining carbon, nitrogen, oxygen, and also some barium abundances are described. The study of these abundances in such stars serves to investigate the means by which the Universe became enriched in metals. Due to the different kinds of CEMP stars observed in the Milky Way, it can only be assumed that there is certainly more than one method of carbon-enhancement at early times. Complete abundance analyses for as many of these archaeological relics as possible are needed in order to constrain the astrophysical sites of early carbon production. There are three main parts of this dissertation. The first part describes new techniques to determine oxygen abundances from spectra of the near-infrared molecular CO bands. With the near-ir OSIRIS spectrograph on the SOAR 4.1-m telescope, 57 CEMP stars were observed. A wide range of oxygen abundances were estimated, and the results were statistically compared to high-resolution estimates for both carbonenhanced and carbon-normal metal-poor stars. Abundance patterns of the sample stars were compared to yield predictions for very metal-poor asymptotic giant branch (AGB) stars. The majority of the sample exhibit patterns consistent with CEMP stars having s-process-element enhancements, and thus have very likely been polluted by carbon- and oxygen-enhanced material transferred from a metal-poor AGB

3 companion. The second part delineates a new survey effort implemented in order to identify new CEMP stars. For the initial pilot study, a new selection technique was developed based solely on the strength of the CH G band at 4300 Å. This technique eliminated previous temperature and metallicity biases present in other CEMP surveys. Observations of the pilot sample were carried out with the Goodman HTS spectrograph on the SOAR 4.1-m telescope. Of the over 120 candidate stars observed, over 35% were found to be CEMP stars. The selection technique was then improved to include a second index for the strength of the G band, and the survey was continued on both the SOAR and Gemini telescopes. After this extension, the success rate of this program increased to 50%. The final part of this dissertation contains details of a pilot study of known metalpoor stars using the X-Shooter spectrograph on the Very Large Telescope (VLT). With three spectrograph arms (near-ultraviolet, optical, and near-infrared), this instrument was used to calculate carbon, nitrogen, oxygen, and barium abundances for a sample of 27 CEMP stars. The broad spectral range of this instrument is unprecedented, and it is an efficient way to estimate abundances for several pertinent species in just one exposure per star. Of the 27 stars, many proved to be enhanced in carbon. The majority appear to be consistent with s-process-element enhancement, but there was one extremely metal-poor star which falls into the rare family of CEMP stars with no neutron-capture-element enhancement.

4 Copyright by CATHERINE R. KENNEDY 2011

5 To my mother and father for the love and support. To Lauren, Nora, Liz, Amanda, Jennie, Cait, Val, Bill, and Dave for the lasting friendship and the laughter. To the person I have looked up to thoughout my entire life, John. I have always endeavored to be his equal with regard to integrity, work ethic, and academic success. He is as fast now as he was on his Big Wheel; but if I pedal fast, I may catch up to him. v

6 ACKNOWLEDGMENTS Many thanks to Timothy C. Beers for his guidance for the past five years. Special thanks to Silvia Rossi, Vinicius Placco, Thirupathi Sivarani, and Young Sun Lee for endless amounts of collaborative spirit and scientific expertise. C.R.K. acknowledges partial support for this work from grants AST , as well as from PHY and PHY ; Physics Frontier Center/Joint Institute for Nuclear Astrophysics (JINA), awarded by the US National Science Foundation. Thanks to Christopher Waters for the LATEX class used to format this thesis. vi

7 TABLE OF CONTENTS List of Tables... List of Figures... ix xi List of Symbols... xv 1 Introduction The Milky Way Components and Formation Scenarios The Carbon-Enhanced Metal-Poor Stars of the Galactic Halo. 3 2 Oxygen Abundances for CEMP Stars from Near-Infrared Spectroscopy with SOAR/OSIRIS Introduction Observations and Data Reduction Adopted Atmospheric Parameters and Synthetic Spectra Determination of [O/H] Results Statistical Comparison to High-resolution Oxygen Estimates High-resolution Nitrogen Estimates Discussion [O/Fe] in Carbon-normal and Carbon-enhanced Metal-poor Stars C, N, and O: Comparison with AGB Models Considering the Effects of Dilution Uncertainties of the AGB Models Conclusions A Survey to Identify New CEMP Stars Motivation The Selection Technique: A New Extended Index for the CH G Band Target Selection Observations and Data Reduction Atmospheric Parameters and Carbon Abundances Results Conclusions vii

8 4 Improving the Selection Technique and Extending the Survey with Multiple Observatories Refining the Selection Criteria New Selection Follow-up Observations: GMOS on Gemini Follow-up Observations: Goodman HTS on SOAR Carbon, Nitrogen, Oxygen, and Barium Abundances with XSHOOTER on VLT Introduction Selected Targets and Observations Model Atmospheres, Line Lists, and Synthetic Spectra Model Atmospheres Line Lists Synthetic Spectra Abundance Determinations Results Discussion Neutron-capture-enhanced Stars CEMP-no Stars Conclusions Conclusions A XSHOOTER Spectra: [C/Fe], [N/Fe], [O/Fe], and [Ba/Fe] Determinations A.1 [C/Fe] A.2 [N/Fe] A.3 [O/Fe] A.4 [Ba/Fe] B Acronyms References viii

9 LIST OF TABLES 1.1 Classification of CEMP Stars Atmospheric Parameters and Carbon Abundances I Atmospheric Parameters and Carbon Abundances II Atmospheric Parameters and Carbon Abundances III Abundance Ratios for the Entire Sample I Abundance Ratios for the Entire Sample II Abundance Ratios for the Entire Sample III Classification of CEMP Candidates Color, GPE, and Category of Pilot Stars I Color, GPE, and Category of Pilot Stars II Color, GPE, and Category of Pilot Stars III Color, GPE, and Category of Pilot Stars IV Color, GPE, and Category of Pilot Stars V Color, GPE, and Category of Pilot Stars VI Atmospheric Parameters and Carbon Abundances for Pilot Stars I Atmospheric Parameters and Carbon Abundances for Pilot Stars II Atmospheric Parameters and Carbon Abundances for Pilot Stars III Atmospheric Parameters and Carbon Abundances for Pilot Stars IV Atmospheric Parameters and Carbon Abundances for Pilot Stars V Atmospheric Parameters and Carbon Abundances for Pilot Stars VI Atmospheric Parameters and Carbon Abundances for Gemini Stars I.. 63 ix

10 4.2 Atmospheric Parameters and Carbon Abundances for Gemini Stars II Atmospheric Parameters and Carbon Abundances for Gemini Stars III Atmospheric Parameters and Carbon Abundances for Goodman Stars I Atmospheric Parameters and Carbon Abundances for Goodman Stars II Atmospheric Parameters and Carbon Abundances for Goodman Stars III Atmospheric Parameters and Barium Abundances for X-Shooter Stars I Atmospheric Parameters and Barium Abundances for X-Shooter Stars II Nitrogen, Carbon, and Oxygen Abundances for X-Shooter Stars I Nitrogen, Carbon, and Oxygen Abundances for X-Shooter Stars II x

11 LIST OF FIGURES 1.1 Path of s- and r-process Three synthetic spectra with different [O/Fe] ratios Estimates of [O/Fe] for four stars Distribution of [O/Fe] [O/Fe] vs. [Fe/H] for stars with [C/Fe] [O/Fe] compared to high-resolution analysis C, N, and O abundances vs. metallicity for 10 stars Abundances compared to AGB abundance yields The effects of dilution on abundances The new GPE index GPE vs. (J K) Index-color distribution of CEMP candidates of different types Metallicity cutoff for CEMP candidates Goodman spectra of pilot-program stars I Goodman spectra of pilot-program stars II Goodman spectra of pilot-program stars III [C/Fe] estimate for pilot-program star [C/Fe] versus [Fe/H] for Pilot Sample The new EGP index for 6 HES stars Distribution of GPE and EGP indices for both the stars and the bright subsets xi

12 4.3 Distribution of GPE and EGP indices after the saturation correction Selection criteria for GPE and EGP indices [C/Fe] versus [Fe/H] for Pilot Sample + Gemini Sample [C/Fe] versus [Fe/H] for Pilot + Gemini + Additional Goodman samples C, N, O, and Ba spectral synthesis C, N, O, and Ba abundances Abundance patterns for CEMP X-Shooter stars CNO Abundances of metal-poor AGB Stars A.1 Carbon abundances for X-Shooter stars A.2 Carbon abundances for X-Shooter stars A.3 Carbon abundances for X-Shooter stars A.4 Carbon abundances for X-Shooter stars A.5 Carbon abundances for X-Shooter stars A.6 Carbon abundances for X-Shooter stars A.7 Carbon abundances for X-Shooter stars A.8 Carbon abundances for X-Shooter stars A.9 Carbon abundances for X-Shooter stars A.10 Carbon abundances for X-Shooter stars A.11 Carbon abundances for X-Shooter stars A.12 Carbon abundances for X-Shooter stars A.13 Carbon abundances for X-Shooter stars A.14 Carbon abundances for X-Shooter stars A.15 Nitrogen abundances for X-Shooter stars A.16 Nitrogen abundances for X-Shooter stars xii

13 A.17 Nitrogen abundances for X-Shooter stars A.18 Nitrogen abundances for X-Shooter stars A.19 Nitrogen abundances for X-Shooter stars A.20 Nitrogen abundances for X-Shooter stars A.21 Nitrogen abundances for X-Shooter stars A.22 Nitrogen abundances for X-Shooter stars A.23 Nitrogen abundances for X-Shooter stars A.24 Nitrogen abundances for X-Shooter stars A.25 Nitrogen abundances for X-Shooter stars A.26 Nitrogen abundances for X-Shooter stars A.27 Nitrogen abundances for X-Shooter stars A.28 Nitrogen abundances for X-Shooter stars A.29 Oxygen abundances for X-Shooter stars A.30 Oxygen abundances for X-Shooter stars A.31 Oxygen abundances for X-Shooter stars A.32 Oxygen abundances for X-Shooter stars A.33 Oxygen abundances for X-Shooter stars A.34 Oxygen abundances for X-Shooter stars A.35 Oxygen abundances for X-Shooter stars A.36 Oxygen abundances for X-Shooter stars A.37 Oxygen abundances for X-Shooter stars A.38 Barium abundances for X-Shooter stars A.39 Barium abundances for X-Shooter stars A.40 Barium abundances for X-Shooter stars A.41 Barium abundances for X-Shooter stars xiii

14 A.42 Barium abundances for X-Shooter stars A.43 Barium abundances for X-Shooter stars A.44 Barium abundances for X-Shooter stars A.45 Barium abundances for X-Shooter stars xiv

15 LIST OF SYMBOLS pc unit of distance equal to cm T eff effective temperature g surface gravity M solar mass xv

16 Chapter 1: Introduction There are a variety of methods by which the Universe was polluted with all of the elements that we observe today. Big Bang nucleosynthesis provided the early Universe with only hydrogen, helium, and trace amounts of lithium. All other elements have since been created in stars. As early generations of stars exploded as supernovae, the interstellar medium (ISM) was polluted with the heavy elements that were created within them and during their explosions. Subsequent generations of stars were then born of this more chemically-enriched material. This cycle of the birth and death of generations of stars continues today, and the main result of this process is that the generations of stars become more and more chemically-enriched than their predecessors as time goes on. In order to quantify this process, one can measure the amount of metals in stars via stellar spectroscopy. In astronomical terms, metals are defined as all elements except for hydrogen and helium. In this dissertation, the metallicity of a star is defined as [Fe/H], a logarithmic comparison of the amount of metal in a star compared to that of the Sun. Other abundances will be discussed in terms of the same notation: [A/B] = log(n A /N B ) star log(n A /N B ), (1.1) where N A and N B are the number densities of the elements A and B for the stars. In these terms, a star which has [Fe/H] = 1.0 has a metallicity of 10 times less than solar, a star which has [Fe/H] = 2.0 has a metallicity of 100 times less than solar 1

17 and so on. 1.1 The Milky Way Components and Formation Scenarios The Milky Way consists of several different physical components. The populations of stars that exist in the three main components (the bulge, the disk, and the halo) have different kinematic and chemical properties. The bulge of the galaxy lies directly in the center and rotates with a mean velocity of 75 km/s. The bulge has a mean metallicity of [Fe/H] = +0.25, which is almost twice the solar metallicity. However, the stars in the bulge span a large range in ages and metallicities (from [Fe/H] = 1.50 to +0.50) and therefore have likely originated from different populations of stars (Zoccali et al., 2008). The disk of the Milky Way is composed of several spiral arms and is of course home to the Sun, which is a distance of approximately 8 kpc away from the center of our Galaxy. The disk itself can be subdivided into separate components: the thin disk and the thick disk. The thin disk contains stars which are roughly comparable in composition to the Sun, having metallicities that are typically within ±2 times the solar value. The thick disk is slightly more metal-poor overall, having metallicities that range from 1.6 < [Fe/H] < 0.4. The halo is the site of some of the oldest and most metal-poor stars in the Milky Way. The focus of this dissertation is primarily on these halo stars. Like the disk of the Galaxy, the halo can also be separated into different components which have different characteristics. The inner halo reaches to kpc from the Galactic center. It is slightly oblate, and the stars that populate the inner halo typically have high orbital eccentricities. The mean metallicity of the inner halo is [Fe/H] = 1.6. There is also a counter-rotating outer halo component (at distances of above 15 kpc 2

18 from the Galactic center) which is more spherical and has a mean metallicity of [Fe/H] = 2.2 (Carollo et al., 2007). The different kinematic properties and chemical properties of the two halo components are indicative of different structure formation episodes of the Galactic halo as a whole. According to current theory, the inner halo was most likely formed by the following hierarchical scenario. A merger of several sub-galactic-mass fragments occurred to form the bulk component of the Milky Way. Subsequent merging of other fragments resulted in the formation of the inner halo which then necessarily contained the higheccentricity stars that are observed today. As the disk of the Galaxy grew in size, the surrounding inner halo component was flattened, giving rise to its now-observed oblate shape (Bekki & Chiba, 2001; Chiba & Beers, 2001). As the outer halo is both more metal poor and counter rotating, it is likely to have been formed by another scenario altogether. It is believed that the outer halo was formed from the accretion of much smaller subsystems, such as the ultra-faint dwarf galaxies that are observed today. Upon such accretion, these systems would suffer severe tidal stripping which would result in the field halo stars that are currently present in the outer halo. Indeed, the recently-studied ultra-faint dwarf galaxies exhibit the very low metallicities and abundance patterns similar to outer-halo field stars (Norris et al., 2010a,b). 1.2 The Carbon-Enhanced Metal-Poor Stars of the Galactic Halo Metal-poor stars in the Milky Way are archaeological relics of the Galaxy s formation. By studying the composition of these objects, it is possible to unearth clues to the formation and nucleosynthetic history of the Milky Way. Large-scale survey efforts such as the HK survey (Beers et al., 1985, 1992) and the Hamburg/ESO Survey (HES; 3

19 Christlieb, 2003; Christlieb et al., 2008) have resulted in the identification of very large numbers of metal-poor stars (with [Fe/H] < 1.0). A significant fraction of metalpoor stars are found to be enhanced in carbon (Lucatello et al., 2006; Marsteller et al., 2009), and the fraction of carbon-enhanced metal-poor stars increases with declining metallicity (Beers & Christlieb, 2005; Frebel et al., 2005; Norris et al., 2007). Much can be learned from the stellar spectra of carbon-enhanced metal-poor stars. Carbon-enhanced metal-poor (CEMP) stars preserve chemical signatures of the Galactic past. By studying their abundance patterns, one can begin to uncover details of the origins of the elements. There exist a number of different types of CEMP stars which have unique abundance characteristics. The different classes, as defined by Beers & Christlieb (2005), are listed in Table 1.1. Table 1.1: Classification of CEMP Stars Class Abundance Pattern CEMP [C/Fe] > +1.0 CEMP-r [C/Fe] > +1.0 and [Eu/Fe] > +1.0 CEMP-s [C/Fe] > +1.0, [Ba/Fe] > +1.0, and [Ba/Eu] > +0.5 CEMP-r/s [C/Fe] > +1.0 and 0.0 < [Ba/Fe] < +0.5 CEMP-no [C/Fe] > +1.0 and [Ba/Fe] < 0.0 The different classes of CEMP stars are suggestive of different sites of carbon production at early times. Often, CEMP stars exhibit enhancements of neutron-capture elements. The most common types of carbon-enhanced metal-poor stars observed are the CEMP-s stars, which show evidence of s-process-element enhancements. It is widely believed that these objects are the result of mass transfer from a companion low-metallicity asymptotic giant branch (AGB) star, where the production of carbon 4

20 and s-process-elements occurs. Stars which have r-process enhancements (CEMP-r) might have evolved from the interstellar medium (ISM) that was polluted by early Type II supernovae where the r-process is believed to have occurred. There are even some CEMP stars which have both s- and r-process element enhancements, suggesting that the source of their carbon-enhancements arises from more than one site and/or time period. While there are no true s- or r-process elements, there are certain isotopes of elements which are exclusive to each process. Figure 1.1 shows a portion of the path of the s- and r-processes. Also in Figure 1.1 is the relative fraction of each element that arises from either process. Figure 1.1: Part of the paths of the s- and r-processes. The columns on the right designate the percentage of each element which is populated by the occurrence of each process. This figure is taken from Sneden et al. (2008). For interpretation of the references to color in this and all other figures, the reader is referred to the electronic version of this dissertation. 5

21 For the CEMP-no class, which exhibit NO neutron-capture element enhancements, the source of carbon enhancement is less certain. Current theories include the possibility that very massive, rapidly-rotating, mega metal-poor stars (with [Fe/H] < 6.0) were very efficient producers of carbon, nitrogen, and oxygen (Hirschi et al., 2006; Meynet et al., 2006). Yet another possible source of CNO enhancement for the observed CEMP-no stars are faint supernovae which undergo heavy mixing and fallback during their explosions (Umeda & Nomoto, 2003, 2005; Tominaga et al., 2007). The majority of CEMP-no stars tend to be some of the most metal-poor objects observed. In fact, they are likely the preservers of the earliest form of carbon production that occurred in the first generations of stars. This idea is confirmed by the recent discovery of a carbon-enhanced, extremely metal-poor damped Ly-α system at a redshift of z = 2.34 (Cooke et al., 2011). The abundances determined by this study are consistent with observed CEMP-no stars in the Galactic halo. By estimating carbon, nitrogen, oxygen, and neutron-capture-element abundances, one can classify CEMP stars and begin to constrain the properties of Galactic Chemical Evolution (GCE). Constraints can also be applied to the Initial Mass Function (IMF) of the Galaxy, as information is revealed about the distribution of different progenitors of CEMP stars and their masses. The organization of this dissertation is as follows. Chapter 2 is a version of the published paper of Kennedy et al. (2011, reproduced by permission of the AAS), and reports new oxygen abundances for known CEMP stars. Chapters 3 and 4 describe a new, highly-successful technique to discover previously-unidentified Galactic CEMP stars. The determinations of C, N, O, and Ba abundances for CEMP stars with the new X-Shooter spectrograph are presented in chapter 5, and the conclusions related to all of these studies are discussed in chapter 6. 6

22 Chapter 2: Oxygen Abundances for CEMP Stars from Near-Infrared Spectroscopy with SOAR/OSIRIS 2.1 Introduction Carbon-enhanced metal-poor (CEMP) stars are quite common in the halo populations of the Milky Way, and are of particular interest, as they preserve important astrophysical information concerning the early chemical evolution of the Galaxy (Beers & Christlieb, 2005). Previous work has indicated that at least 20% of stars with metallicities [Fe/H] < 2.0 exhibit large overabundances of carbon ([C/Fe] > +1.0; Lucatello et al., 2006; Marsteller et al., 2009), although recent studies (e.g. Cohen et al., 2005; Frebel et al., 2006), have claimed that this fraction is somewhat lower (9% and 14%, respectively, for [Fe/H] < 2.0). In any case, the fraction of CEMP stars rises to 30% for [Fe/H] < 3.0, 40% for [Fe/H] < 3.5, and 100% for [Fe/H] < 4.0 (Beers & Christlieb, 2005; Frebel et al., 2005; Norris et al., 2007). There exist a number of classes of CEMP stars, some of which have been associated with proposed progenitor objects. CEMP-s stars (those with s-process-element enhancement), for example, are the most commonly observed type to date. Highresolution spectroscopic studies have revealed that around 80% of CEMP stars exhibit s-process-element enhancement (Aoki et al., 2007). The favored mechanism 7

23 invoked to account for these stars is mass transfer of carbon-enhanced material from the envelope of an asymptotic giant branch (AGB) star to its binary companion; it is this surviving binary companion that is now observed as a CEMP-s star. The class of CEMP-no stars (which exhibit no strong neutron-capture-element enhancements) is particularly prevalent among the most metal-poor stars. Possible progenitors for this class include massive, rapidly rotating, mega metal-poor ([Fe/H] < 6.0) stars, which models suggest have greatly enhanced abundances of CNO due to distinctive internal burning and mixing episodes, followed by strong mass loss (Meynet et al., 2006; Hirschi et al., 2006; Meynet et al., 2010). Another suggested mechanism is pollution of the interstellar medium by the so-called faint supernovae associated with the first generations of stars, which experience extensive mixing and fallback during their explosions (Umeda & Nomoto, 2003, 2005; Tominaga et al., 2007); high [C/Fe] and [O/Fe] ratios are predicted in the ejected material. This model well reproduces the observed abundance pattern of the CEMP-no star BD+44:493, the ninth-magnitude [Fe/H] = 3.7 star (with [C/Fe]= +1.3, [N/Fe] = +0.3, and [O/Fe] = +1.6) discussed by Ito et al. (2009). The great majority of known CEMP stars were originally identified as metal-poor candidates from objective-prism surveys, such as the HK Survey (Beers et al., 1985, 1992), and the Hamburg/ESO Survey (HES; Christlieb, 2003; Christlieb et al., 2008), based on a weak (or absent) Ca II K line. Some candidate CEMP stars also come from a list of HES stars selected from the prism plates based on their strong molecular lines of carbon (Christlieb et al., 2001). Medium-resolution spectra for most of these objects have been obtained over the past few years (Goswami et al., 2006, 2010; Marsteller, 2007, T. Sivarani et al. 2011, in preparation). Inspection of these data indicates that at least 50% of these targets are consistent with identification as CEMP stars, while the others are roughly solar-metallicity carbon-rich stars. Dedicated surveys for CEMP stars covering a wide range of carbon abundance and metallicities 8

24 are just now getting underway, based on the observed strength of the CH G band measured from the HES prism plates (e.g., Placco et al., 2010). In order to more fully test the association of CEMP-no stars with massive primordial stars and/or faint supernovae, and to better explore the nature of the s-process in low-metallicity AGB stars (which is still rather poorly understood; Herwig, 2005), we require measurements of the important elements C, N, and O for as large a sample of CEMP stars as possible. While estimates of carbon and nitrogen abundances can be determined from medium-resolution optical or near-ultraviolet spectra of CEMP stars (e.g., Rossi et al., 2005; Beers et al., 2007b; Johnson et al., 2007; Marsteller et al., 2009), high-resolution spectroscopy is usually required in order to obtain estimates of oxygen abundances from the forbidden [O1] λ6300 line, the λ7700 triplet (e.g., Schuler et al., 2006; Sivarani et al., 2006; Fabbian et al., 2009, and references therein), or the OH lines at µm (Meléndez & Barbuy, 2002). Masseron et al. (2010) provide a useful compilation of known elemental abundances for CEMP stars. In addition to abundance measurements for metal-poor halo stars, oxygen abundances have also been measured directly in the gas phase in damped Lyα systems (Pettini et al., 2002, 2008). If a star has a measured carbon abundance (and, assuming C/O > 1, which applies for most CEMP stars), essentially all of the O is locked up in CO molecules, and medium-resolution spectroscopy of the CO ro-vibrational bands in the near-infrared (near-ir) can be used for the estimation of [O/Fe] (e.g., Beers et al., 2007b, and references therein). Although one sacrifices measurement accuracy, relative to highresolution studies, this approach has the great advantage that medium-resolution spectroscopy can be gathered far faster than high-resolution spectroscopy, ensuring that much larger samples of stars can be investigated. In addition, the large separation of the 13 CO lines from the 12 CO lines at 2.3µm provides a straightforward means to measure the important mixing diagnostic 12 C/ 13 C, as long as the signal-to-noise ratio 9

25 (S/N) of the spectra is sufficient. This paper is outlined as follows. In Section 2.2, we discuss details of the observations and data reduction procedures used in the present study. Section 2.3 describes the previously determined atmospheric parameter estimates and their origins, as well as details about the synthetic spectra. Methods used for the determination of [O/Fe] for our sample of stars are described in Section 2.4. Our results and a statistical comparison to high-resolution estimates of [C/Fe] and [O/Fe] for a subset of our program stars can be found in Section 2.5. Section 2.6 is a short discussion of our results; conclusions follow in Section Observations and Data Reduction Our sample of 57 stars was selected from the HES, based on follow-up mediumresolution optical spectra obtained during the course of searches for low-metallicity stars. These optical spectra were obtained with the GOLDCAM spectrograph on the KPNO 2.1m telescope and with the RC Spectrographs on the 4m KPNO and CTIO telescopes (see Beers et al., 2007b, hereafter Paper I). Additional targets were selected from the list of carbon-rich candidates published by Christlieb et al. (2001) with available optical spectra. Based on the optical spectra, all of the candidates are metal-poor stars, spanning the metallicity range 2.8 [Fe/H] 1.0. All of the stars were selected to be carbon-rich, with the majority exhibiting [C/Fe] +1.0, and thus are carbon enhanced as defined by Beers & Christlieb (2005). Since our intention was to obtain near-ir spectroscopy of the CO features, the stars were also selected to have effective temperatures less than 5000 K, since warmer stars do not exhibit strong CO. Estimates of [O/Fe] for our program stars are derived from the analysis of mediumresolution near-ir spectra taken with the SOAR 4.1m telescope, using the OSIRIS (Ohio State Infrared Imager/Spectrometer; Depoy et al., 1993) spectrograph during 10

26 2005 October to 2008 June. We used the long slit (width set to 1 ) and long camera (with focal ratio f/7), which provided a resolving power R = The long-pass K-band filter was used to isolate the spectral region from 2.25µm to 2.45µm. Visible in this band are the four ro-vibrational CO features used for the determination of [O/Fe]. We also observed A0-type stars at the same air mass as the observations of the program objects in order to correct for the presence of telluric lines in the spectra. The K-band magnitude range for our sample stars is 7 12, resulting in exposure times in the range seconds in order to reach our targeted S/N of 50/1. Spectra of Ar Ne arc lamps, taken before or after each program star, were used for the wavelength calibration of our sample. Bias correction, flat-fielding, spectral extraction, wavelength calibration, telluric feature correction, and continuum normalization were all performed using standard IRAF packages. IRAF is distributed by the National Optical Astronomy Observatory, which is operated by the Association of Universities for Research in Astronomy, Inc., under cooperative agreement with the National Science Foundation. 2.3 Adopted Atmospheric Parameters and Synthetic Spectra Atmospheric parameters (T eff, log g, and [Fe/H]) were estimated from available optical and near-ir photometry as well as from previously obtained medium-resolution optical spectroscopy. Estimates of T eff are obtained from measured V K colors (taken from Beers et al., 2007a, and references therein, as well as from the 2MASS Point Source Catalog; Skrutskie et al. 2006). The use of near-ir photometry provides for a more accurate determination of T eff, as the K band is less influenced by the presence of carbon features than bluer bands. We used the Alonso et al. (1996) calibrations of T eff with V K colors, as described in Paper I. Surface gravities, 11

27 log g, have been estimated based on the Padova evolutionary tracks for metallicities [Fe/H]= 2.5 and [Fe/H]= 1.7 (Girardi et al., 2000; Marigo et al., 2001). Uncertainties in T eff and log g are 100 K and 0.5 dex, respectively. The microturbulence is taken to be 2 km s 1 for all stars. This is consistent with previously determined microturbulence values for giant CEMP stars (Johnson et al., 2007; Aoki et al., 2007). We have constructed two sets of synthetic spectral templates, covering the optical and near-ir bands. Each set consists of 2000 synthetic spectra with carbon-enhanced atmospheres generated with the MARCS code (Gustafsson et al., 2008). We used a previous generation of models here, as updated CEMP models were not available. We do not, however, anticipate large differences in the models of spectra. The use of carbon-enhanced models is of particular importance for cool CEMP stars, for which the atmospheric structure is significantly altered by carbon (Masseron et al., 2006). No 3D 1D corrections have been applied to our estimates. Recent studies of these effects on two hyper metal-poor stars (Collet et al., 2006) have revealed [O/Fe] corrections of 0.8 based on OH molecules, thereby lowering the measured abundance of oxygen. However, the magnitude of such corrections is expected to decrease with increasing metallicity (Collet et al., 2007). As the metallicities of our targets range from 1.0 to 2.8, it is likely that the three-dimensional effects are less severe. In addition, more recent studies have been carried out (A. Ivanauskas, private communication) concerning the effects of convection on C 2, CH, CN, CO, NH, and OH molecules. When compared to the results from Collet et al. (2006), the magnitude of the corrections appears smaller. The synthetic grid covers a range T eff = 4000 to 6000 K, log g = 0.0 to 5.0, [Fe/H] = 5.0 to 0.0, and [C/Fe] = 0.0 to We adopt fixed nitrogen abundances set to 0.5 dex less than the carbon abundances, which is roughly appropriate for CEMP stars. The CH and CN line lists used for the synthesis of the optical spectra are those compiled by Plez (see Plez & Cohen, 2005). The CO line lists used for the near-ir 12

28 synthesis are taken from Kurucz (1993). The synthetic grids are then degraded to match the resolving power of the observed spectra (R = 2000 for the optical spectra and R = 3000 for the near-ir spectra). The optical spectra are used for the determination of [Fe/H] and [C/H]. The Ca II K line is matched with the model spectra to estimate [Fe/H], and the C 2 and CN features are fit for the estimation of [C/H] (see Paper I). Our adopted atmospheric parameters, as well as the derived [C/H] and [C/Fe], are listed in Tables 2.1, 2.2, and 2.3. Table 2.1: Atmospheric Parameters and Carbon Abundances I Name Teff (K) log g (cgs) [Fe/H] [C/H] [C/Fe] HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE

29 Table 2.2: Atmospheric Parameters and Carbon Abundances II Name Teff (K) log g (cgs) [Fe/H] [C/H] [C/Fe] HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE

30 Table 2.3: Atmospheric Parameters and Carbon Abundances III Name Teff (K) log g (cgs) [Fe/H] [C/H] [C/Fe] HE HE HE HE HE HE HE HE HE HE HE HE HE HE Determination of [O/H] In order to determine [O/H], we employed the near-ir synthetic spectra constructed from model atmospheres with carbon enhancements (see above). Each synthetic spectrum covers the wavelength range 2.25µm 2.40µm. For all combinations of these parameters, models were available with [O/Fe] values of 0.0, +0.4, and The first step in the determination of [O/Fe] is to use the grid of synthetic spectra in combination with the atmospheric parameters to create three models with [O/Fe] 15

31 values of 0.0, +0.4, and These models are then used in order to estimate the [O/Fe] of the program stars. Each star in the sample has a previously estimated T eff, log g, [Fe/H], and [C/Fe]. Given the fact that four parameters are known, the routine begins with 48 models: 16 models for each value of [O/Fe]. The 48 models are selected as having the two closest values of T eff, log g, [Fe/H], and [C/Fe] for each of the three values of [O/Fe]. Once selected, a linear interpolation over each parameter is performed in order to create three final models, one for each value of [O/Fe], with the known values of the four parameters. The three final models for a set of typical atmospheric parameters are shown in Figure 2.1. With the other parameters fixed, it is easy to see how a typical spectrum changes with increasing oxygen abundance. Next, a linear interpolation over [O/Fe] is performed on the three final models, creating new model spectra with varying oxygen. It should be noted that while the interpolation was performed over [O/Fe], the metallicity is fixed, and therefore the oxygen varies as [O/H]. In some cases, it was necessary to extrapolate beyond the boundaries of the model grid in order to find a good fit to the data. Due to the presence of four large CO bands in the near-ir spectra, there were difficulties in fitting the continuum across the entire region. It was important to fit the continua with a low-order function so that the depth of the absorption features was not artificially enhanced or lessened due to the continuum fit. For this reason, the spectra of all stars in the sample were trimmed around each of four CO bands, and a local continuum was fit for each band prior to spectral synthesis. With the use of the synthetic spectra, oxygen abundances were then estimated individually for each of the four bands by minimizing χ 2. A robust average using bisquare weighting of the four separate values was taken as the final estimate of oxygen abundance, with an associated robust estimate of the scatter in these values taken as the error of determination. Figure 2.2 shows the fitting technique applied to four stars from the sample. Each 16

32 Figure 2.1: Three synthetic spectra with different [O/Fe] ratios. Each spectrum has T eff of 4500 K, log g of 1.0, [Fe/H] of 2.0, and [C/Fe] of row shows the four separate estimates of [O/H] for each star. Also plotted on each panel are synthetic spectra with [O/H] values that vary from the best-fitting spectra by ±0.5 dex. Once the robust average is applied the resulting estimates of [O/H] are 1.2 for HE , 0.5 for HE , 0.6 for HE , and 1.2 for HE Results The distributions of [O/Fe] versus three of the parameters used for their determinations are shown in Figure 2.3. The solid lines are linear fits to the data. For the entire sample, there are no significant correlations of [O/Fe] with T eff, [Fe/H], or [C/Fe]. In the middle panel of Figure 2.3, it can be seen that only one of the stars in our sample with [Fe/H] < 2.5 has a value of [O/Fe] less than The bottom panel of Figure 17

33 Figure 2.2: Each row shows four estimates of [O/H] for a star from our sample: HE , HE , HE , and HE , respectively. In each panel, the black lines are the data, the red lines are the best-fitting synthetic spectra, the green lines have [O/H] values of 0.5 dex lower than the best-fitting spectra, and the blue lines have [O/H] values of 0.5 dex higher than the best-fitting spectra. A robust average of the four separate estimates is taken as the final estimate of oxygen abundance for each star. See the text for details. 2.3 shows the distribution of [O/Fe] versus [C/Fe], revealing that the majority of our sample (45 stars) have [C/Fe] > +1.0, and so meet the definition for CEMP stars given by Beers & Christlieb (2005). The other stars exhibit carbon enhancements of [C/Fe] +0.5 and thus are at least moderately enhanced in carbon. The average error in the determination of [O/Fe] for our entire sample is 0.4 dex. We adopted a minimum error for our [O/Fe] estimates of 0.25 dex due to the influence of errors that arise from the estimation of T eff, log g, [Fe/H], and [C/Fe] (see Paper I for details). In Figure 2.4, a carbon cut has been made, such that only the oxygen abundances 18

34 Figure 2.3: Top panel: [O/Fe] vs. T eff for the entire sample. Middle panel: [O/Fe] vs. [Fe/H] for the entire sample. Bottom panel: [O/Fe] vs. [C/Fe] for the entire sample. for those stars with [C/Fe] > are plotted against [Fe/H]. The stars with the highest abundances of carbon exhibit some of the lowest metallicities in our sample. This is not surprising, given that high values of [C/Fe] are often associated with lower metallicities. The fit to the data shows a slight trend of increasing oxygen with decreasing metallicity. The solid line is a least squares fit of [O/Fe] as a function of [Fe/H]. Only a marginally significant slope ( ± 0.314) is found, hence the correlation is quite weak. For comparison, the dashed line in this figure represents the fit of [O/Fe] versus [Fe/H] for the carbon-normal stars from the Spite et al. (2005) sample. The [O/Fe] estimates of the Spite et al. (2005) sample come from the forbidden O I λ6300 line. 19

35 Figure 2.4: [O/Fe] vs. [Fe/H] for the stars with [C/Fe] The dashed line represents the fit for the carbon-normal stars from the Spite et al. (2005) sample of very metal-poor stars, while the solid line is the best fit for our data Statistical Comparison to High-resolution Oxygen Estimates The present catalog of measured oxygen abundances available in the literature for CEMP stars is still relatively small, due to the difficulty of obtaining estimates from optical spectra, even at high spectral resolution. However, we can at least compare the regions of the [O/Fe] versus [Fe/H] parameter space that are occupied by CEMP stars of various sub-classes, based on previous high-resolution oxygen estimates, with those from our present medium-resolution effort. Figure 2.5 shows [O/Fe] for our entire sample, with different boxes indicating the regions of parameter space occupied by several classes of CEMP stars. Sources for 20

36 the high-resolution data for different classes of CEMP stars can be found in Masseron et al. (2010), and references therein. The majority of the stars in our sample occupy regions of the diagram as CEMP stars that have confirmed, high-resolution measurements of s-process-element enhancements. There is clearly overlap with the region occupied by CEMP-r/s stars as well. Few of our stars overlap with the region occupied by CEMP-no stars in the literature; CEMP-no stars tend to be more metal-deficient than most of the stars in our sample. Figure 2.5: [O/Fe] vs. [Fe/H] for the entire sample. The colored boxes show the regions occupied by different types of CEMP stars found in Masseron et al. (2010) High-resolution Nitrogen Estimates For 13 of our program stars, high-resolution estimates of [N/Fe] are available from S. Lucatello (private communication) and/or Aoki et al. (2007). We selected 10 of these 21

37 stars for which the available high-resolution estimates of [C/Fe] were within 0.5 dex of our medium-resolution estimates. The three that are omitted from our analysis and discussion have associated high-resolution [Fe/H] and/or [C/H] estimates that differ significantly from the medium-resolution estimates of these species. We report values of high-resolution [C/Fe], [C/Fe] h, using our estimates of [Fe/H] combined with the high-resolution [C/H]. We report high-resolution estimates of [N/Fe], [N/Fe] h, by combining our estimates of [Fe/H] with the high-resolution values of [N/H]. Two of the 10 stars had high-resolution estimates both from Lucatello and Aoki et al. (2007), and an average of the two available estimates was taken. The values of [C/Fe], [C/Fe] h, [N/Fe] h, [O/Fe], and σ [O/Fe] for our entire sample are listed in Tables 2.4, 2.5, and 2.6. Table 2.4: Abundance Ratios for the Entire Sample I Name [C/Fe] [C/Fe]h [N/Fe]h [O/H] [O/Fe] σ([o/f e]) HE HE HE HE HE HE HE HE HE HE HE HE HE HE

38 Table 2.5: Abundance Ratios for the Entire Sample II Name [C/Fe] [C/Fe]h [N/Fe]h [O/H] [O/Fe] σ([o/f e]) HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE

39 Table 2.6: Abundance Ratios for the Entire Sample III Name [C/Fe] [C/Fe]h [N/Fe]h [O/H] [O/Fe] σ([o/f e]) HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE Discussion We expect that the majority of CEMP stars in our sample have been polluted by a companion low-metallicity AGB star. In an AGB star, intershell oxygen is predicted 24