EDP Sciences logo
Subscriber Authentication Point
Search
Menu
Home All issues Volume 593 (September 2016) A&A, 593 (2016) A28 Full HTML
Free Access
Issue
A&A
Volume 593, September 2016
Article Number A28
Number of page(s) 9
Section Stellar atmospheres
DOI https://doi.org/10.1051/0004-6361/201628847
Published online 01 September 2016

© ESO 2016

1. Introduction

Very metal-poor (VMP; [Fe/H] < −2.0) and extremely metal-poor (EMP; [Fe/H] < −3.0) stars provide the opportunity of deepening our understanding of the early chemical evolution of the Milky Way and the Universe. In most cases, their atmospheres exhibit the chemical compositions of the gas from which they formed, enriched by the nucleosynthetic yields of the first stellar populations. In addition, recent investigations of the Milky Way’s stellar halo have revealed that it is not a homogeneous entity, but instead comprises multiple populations (Carollo et al. 2007, 2010; de Jong et al. 2010; Nissen & Schuster 2010, 2011; Beers et al. 2012; An et al. 2013, 2015; Allende Prieto et al. 2014; Chen et al. 2014; Janesh et al. 2016). VMP and EMP stars are therefore useful probes of the assembly of the Galaxy as well.

Early survey work, for example, the HK survey of Beers and colleagues (Beers et al. 1985, 1992) and the Hamburg/ESO survey of Christlieb and collaborators (Reimers & Wisotzki 1997; Christlieb 2003), provided the first large lists of several thousand VMP and EMP stars. More recently, advances in astronomical instrumentation have enabled multiplexed spectroscopy of even larger numbers of (generally fainter) stars, such as the Sloan Digital Sky Survey (SDSS; York et al. 2000); this produced samples of VMP and EMP stars that cover a wide range of distances from the Sun.

High-resolution spectroscopic follow-up of these targets has provided abundance measurements for numerous elements, revealing the existence of chemical peculiarities that have greatly expanded our knowledge of the different nucleosynthetic pathways that contributed to the early chemical evolution of the Galaxy (e.g., Cayrel et al. 2004; Arnone et al. 2005; Aoki et al. 2005, 2013; Cohen et al. 2004, 2007, 2008; Bonifacio et al. 2009; Lai et al. 2009; Roederer 2009; Roederer et al. 2014). Until recently, relatively few stars had confirmed metallicities of [Fe/H] < −3.0, making it difficult to infer the global characteristics of EMP stars. Since 2005, more than 200 stars with [Fe/H] < −3.0 have been confirmed based on high-resolution spectroscopic analyses (e.g., Barklem et al. 2005; Caffau et al. 2013a,b; Yong et al. 2013), including more than 50 stars with [Fe/H] < −3.5, and over 20 with [Fe/H] < −4.0 (Barklem et al. 2005; Frebel et al. 2005, 2015; Cohen et al. 2008; Caffau et al. 2011a,b, 2012, 2013b; Bonifacio et al. 2012; Aoki et al. 2013; Smolinski et al. 2011; Yong et al. 2013; Keller et al. 2014; Allende Prieto et al. 2015; Li et al. 2015; Placco et al. 2015; Meléndez et al. 2016).

The SDSS database comprises over 900 000 stellar spectra and is now the dominant source of confirmed EMP stars. However, the relatively low resolution (R ~ 2000) and limited signal-to-noise ratio (S/N~30−40) of the SDSS spectra themselves make it difficult to derive elemental abundances at very low metallicities.

This paper is the third in a series devoted to analyses of Milky Way halo stars based on the low-resolution SDSS spectroscopic data. We present 107 stars with metallicities [Fe/H] < −3.0 for which the stellar atmospheric parameters and chemical abundances of [Fe/H], [Mg/H], and [Ca/H] have been estimated. We also report on a method for estimating upper limits for these abundances from the SDSS spectra. This technique enables determining the lowest required S/N for estimating the abundance for a given element as a function of effective temperature (Teff) and surface gravity (log g).

This paper is outlined as follows. In Sect. 2 we briefly discuss the data used in the analysis, which is described in Sect. 3. In this section we also introduce our method for estimating upper limits. Section 4 reports on a comparison with estimates from other analyses, including high-resolution spectroscopy. The results are summarized in Sect. 5. Finally, we present our conclusions and a brief discussion in Sect. 6.

2. Observations

Our stellar spectra come from the SDSS. This project, started in 2000, is now on its third extension, SDSS-IV (Alam et al. 2015), and comprises a set of surveys devoted to a variety of areas, from cosmology to the evolution of galaxies and the Milky Way to the search for extrasolar planets. Its first extension, SDSS-II, included a specific stellar project, the Sloan Extension for Galactic Understanding and Exploration (SEGUE; Yanny et al. 2009), directed at investigating the structure, formation, and chemical evolution of the Galaxy. SEGUE-2, a subsurvey of SDSS-III (Eisenstein et al. 2011), increased the number of stellar spectra and focused on observing distant halo stars.

SEGUE, SEGUE-2, and other SDSS programs (including the main SDSS galaxy redshift survey and BOSS, the Baryon Oscillations Spectroscopic Survey, see Dawson et al. 2013) obtain spectra for color-selected samples of F-type main-sequence turn-off stars observed for calibration purposes. Calibration stars taken during the main survey and BOSS have the advantage of being located across the entire SDSS footprint at high Galactic latitudes (instead of the limited number of directions probed by SEGUE and SEGUE-2) and include halo stars at distances of up to ~100 kpc.

The spectra were obtained with a pair of spectrographs on the SDSS 2.5 m telescope (Gunn et al. 2006; Smee et al. 2013) at Apache Point Observatory, with a wavelength-dependent resolving power of 1300 < R < 3000 for the spectral range ~3800 < λ < 9000 Å. The spectrographs were updated before the beginning of BOSS observations to increase their efficiency and spectral range (to 3600 < λ < 10 000 Å). Additional details on these spectra can be found in the technical papers of the surveys (York et al. 2000; Yanny et al. 2009; Dawson et al. 2013; Alam et al. 2015), as well as in previous papers in this series (Allende Prieto et al. 2014; Fernández-Alvar et al. 2015; hereafter Papers I and II).

3. Analysis

3.1. Measurement of stellar parameters and chemical abundances

thumbnail Fig. 1

Model fits for a SDSS/SEGUE star, SDSS J093339.24+310245.4, with the mean S/N of the sample, ~40. The three panels in the left column correspond to the fit regions from which the [Fe/H] abundance was determined. The derived iron abundance is shown in the legend of the upper panel, while the temperature and surface gravity estimates are shown in the legend of the lower panel. The panels in the right column show the CaII HK doublet and the MgIb triplet that were fit to determine Ca and Mg abundances, with the derived estimates shown in the legends.

We wish to estimate stellar atmospheric parameters and individual element abundances for a sample of extremely metal-poor stars belonging to the halo system. Following the same strategy as described in Papers I and II, we made use of the FERRE code (Allende Prieto et al. 2006) to constrain the stellar atmospheric parameters: effective temperature, Teff, surface gravity, log g, and the global metallicity, [M/H]. We adopted the notation [X/H] =, where X is any chemical element, N(X) is the number density of the nuclei of this element, and N(H) is the number density of hydrogen nuclei. [M/H] is the iron abundance determined from fitting the available spectral range, which includes spectral features from other metals.

The search was performed by comparison with a grid of synthetic spectra covering a wide range of parameter space, seeking the minimum χ2 using quadratic Bezier interpolation between the model spectra (Auer 2003). We employed the same three-dimensional spectral library (Teff, log g, and [M/H]) as in Paper I, calculated from one-dimensional plane-parallel Kurucz model atmospheres (Castelli & Kurucz 2003), which consider local thermodynamical equilibrium (LTE). The grid covers the ranges 4750 < Teff < 6500 K, 0.5 < log g < 4.5, and −5.0< [M/H] <+ 0.5, in steps of 250 K, 0.5 dex, and 0.5 dex, respectively. More details of the model atmospheres, line data, and the opacities used in the generation of the spectral library can be found in Paper I.

The FERRE routine enables searches for one, several, or all of the atmospheric parameters in a library of synthetic spectra. When the atmospheric parameters were determined, we searched for a limited set of elemental abundances, holding the atmospheric parameters fixed from the analysis of the full spectrum. In our model grids we varied the abundances of all metals relative to hydrogen in solar proportions (with the exception of the α-elements, which are enhanced by +0.4 dex for metal-poor stars). Searching for [M/H], but restricting the fit to regions dominated by individual lines of other elements, enables estimating the abundance of each of these elements, corresponding to the value of [M/H] that best reproduces the shapes of their associated lines. A more detailed explanation can be found in Paper II.

We first estimated the stellar parameters using the full spectral range provided by the observations. The spectral range was limited to 3850 < λ < 9190 Å to ensure consistency with our previous analysis in Papers I and II, which examined the BOSS spectra and those from earlier SDSS/SEGUE observations. From this analysis we selected our primary targets to have metallicities in the range −4.0< [M/H] <−3.0. We rejected spectra that appeared to be double-lined binaries or white dwarfs. The S/N (calculated as the median value per pixel in the range 4885 < λ < 5500 Å for each spectrum) varied between 20 < S/N < 90; the derived values of Teff and log g varied between 5170 < Teff < 6500 K and 0.5 < log g < 4.5, respectively. Holding Teff and log g fixed, we repeated the search in the [M/H] dimension of the grid by fitting selected regions in the spectra that contain atomic lines of Fe, Ca, or Mg.

In this analysis we fit the following spectral ranges (shown in Fig. 1): 4885 < λ < 5070 Å, 5220 < λ < 5280 Å, and 5295 < λ < 5500 Å to determine Fe abundances; 5160 < λ < 5190 Å (the Mg Ib triplet) for Mg abundances; and 3910 < λ < 3990 Å (the Ca II H and K resonance doublet) for Ca abundances. These are the regions with the highest sensitivity in the optical spectral range for each element. The spectra were normalized by splitting them into 100 Å pieces (200 Å for the Fe I window between 4875 and 5510 Å) and dividing each piece by its mean flux value. Our synthetic spectra were treated in the same manner. From the new metallicity estimates, we obtained the abundance values of [Mg/H] and [Ca/H] ([Fe/H] is straightforward), considering the relation with [Fe/H] adopted in the construction of the spectral grid.

3.2. Method for estimating upper limits

The relatively low S/N and modest resolution of our data complicate the estimation of some elemental abundances, in particular for Fe and Mg, whose line detections become marginal for low metallicities at the S/N of our spectra. For this reason, we developed a method for determining upper limits on the abundance of an element as a function of S/N, Teff, log g, and metallicity, [M/H].

We simulated the observed spectra by smoothing our spectral library to R = 2000 and adding Gaussian noise. From a simulated spectrum with a particular S/N, Teff, log g, and [M/H] (hereafter [M/H]0), we evaluated the χ2 and its error (), comparing with noise-free spectra over −5.0< [M/H] <−2.5 for the same spectral windows as were used to determine the abundances of Fe, Mg, and Ca. We calculated the slope of the χ2 curve in the range −5.0< [M/H] < [M/H]0. Figure 2 illustrates the method. Repeating this process at different S/Ns, from 20 to 90 (in steps of 5), we defined the minimum S/N (hereafter S/N0) at which the slope becomes significant (i.e., its error is lower than its value) for each [M/H]0. This calculation was performed a hundred times to statistically refine this lowest required S/N0 value.

By evaluating simulations at Teff values from 4750 to 6500 K in steps of 250 K, two log g values, 1.5 and 4.2, and metallicities from −4.8 to −2.5 in steps of 0.1 dex, we established the lowest required S/N0 vs. [Fe/H]0 curves for each combination of stellar atmospheric parameters. The S/N range considered was 20−90 for Fe and Mg. We extended the lower limit to S/N = 5 in the Ca evaluation, since the S/N over 3900 < λ < 4000 Å in the analyzed spectra is lower than 20 in some cases. These curves were used to evaluate the reliability of our Fe, Mg, and Ca abundance determinations. Figure 3 shows the results for log g = 1.5 and 4.2. We accepted a given abundance determination when it was higher than the corresponding S/N at which the modulus of the difference with the S/N of the observation is smallest. Otherwise, when the value determined by FERRE was lower than this limit, we considered the estimate as an upper limit for the abundance.

This method can be used for the analysis of spectra with a low resolution and low S/N. From Fig. 3 we can infer that Teff affects the abundance determination of the three elements more strongly than log g, as expected.

Iron is the most difficult to measure. Although more lines are available than for the other two elements, their weakness in this metallicity regime makes them challenging to measure in the presence of noise. Our calculations indicate that a S/N>90 is necessary to reliably determine [Fe/H] ≤ −4.2 over the Teff and log g ranges considered and our selected spectral range. At S/N~ 90, the upper limit changes by ~1 dex, from −4.0 to −3.0, between 4750 to 6500 K at log g = 1.5 (becoming larger as log g increases). By contrast, the calcium abundance can be measured down to [Ca/H] ~ −4.2 for spectra with a median S/N~ 10 for stars with Teff = 4750 K and log g = 4.2.

thumbnail Fig. 2

χ2 obtained from the evaluation of a simulated spectrum at [M/H]0= [Fe/H]0= −3.8 (marked with a vertical dotted line) compared with synthetic spectra over a range of metallicities [M/H] (from −5.0 to −2.5 with a step of 0.1 dex). The blue line shows the resulting χ2 curve for a simulation with S/N = 30, and the red line applies to S/N = 80. The two black dotted lines indicate the linear fits up to the [M/H]0 corresponding to the minimum χ2, from which the slope and its uncertainty are derived.

thumbnail Fig. 3

Lowest required signal-to-noise values, S/N0, to reliably determine [Fe/H]0, [Mg/H]0 and [Ca/H]0 abundances, evaluated at log g = 1.5 and log g = 4.5 and 4750 < Teff < 6500 K (darker colors correspond to lower temperatures). The increase in Teff complicates the [Fe/H] determination, mainly for Fe and Mg, and a higher S/N is necessary. The sensitivity of the Ca II HK resonance lines enables determining [Ca/H] at very low abundances.

To provide a convenient tool for estimating the lowest required S/N for measuring the abundances of these elements from the spectral windows considered in this work, we derived an analytic function that fits the curves obtained from the simulations. We found that these curves can be well-reproduced by a second-order polynomial (a convex parabola) and a straight line through the vertex of the parabola, We modeled the dependence of the polynomial coefficients with Teff as second-order polynomials. The applicability domain for each second-order polynomial is defined below and above certain Teff values (5500 K at log g = 1.5 and 6000 K at log g = 4.2).

A single first-order polynomial fits the dependence of the other two coefficients on Teff well,

Tables 1–3 show the coefficients of these polynomials for each element for the two log g values evaluated, 1.5 and 4.2.

Table 1

Second-order polynomial coefficients that reproduce the parabolic and linear polynomial parameters as a function of Teff for each of the log g values considered, 1.5 and 4.2, to estimate Fe.

Table 2

Same as in Table 1, but for Mg.

Table 3

Same as in Table 1, but for Ca.

4. Verification with other analyses

4.1. Comparison with the SSPP analysis

The SEGUE stellar parameter pipeline (SSPP) was developed to analyze the stellar spectra gathered in the SDSS/SEGUE surveys (see Lee et al. 2008a,b; Allende Prieto et al. 2008; Smolinski et al. 2011; Lee et al. 2011 for details) and can be used to estimate Teff, log g, [Fe/H], and [α/Fe] – the [α/Fe] corresponds to the [Mg/Fe], [Si/Fe], [Ca/Fe], and [Ti/Fe] global measurement from the spectral range 4500−5500 Å fit to the observational data with synthetic spectra. Recently, the SSPP has been extended to be capable of estimating [C/Fe] as well (Lee et al. 2013). For the sake of comparison, we also evaluated our sample of moderate-resolution SDSS spectra with this tool.

We compared the stellar parameters and abundance measurements obtained with the SSPP with our present analysis results. Our analysis returns a lower estimated Teff than the SSPP (δ = −353 K), and a modest dispersion (σ= 277 K). This is expected because our method is purely spectroscopic, and the SSPP uses a combination of photometric and spectroscopic techniques. It has long been recognized that spectroscopically determined temperature estimates can be up to several hundred Kelvin cooler than photometric estimates. Regarding the log g estimate, a large systematic deviation exists for stars for which we obtain surface gravity estimates of log g < 3 (δ = −1.26 dex, σ= 1.56 dex). This is again not surprising because estimates of surface gravity are particularly challenging at low metallicity from low-S/N spectra. The [Fe/H] comparison exhibits a relatively small negative offset δ = −0.16 dex, with a dispersion σ= 0.33 dex. Essentially all of this offset can be accounted for by the differences in the temperature estimates.

Finally, we compared the SSPP estimate of [α/Fe] with our [Ca/Fe] and [Mg/Fe] measurements. Both show an offset δ ~ −0.15 dex and dispersion σ~ 0.26 dex. Stars with the highest and lowest [Ca/Fe] and [Mg/Fe] estimates from our own estimates exhibit the largest differences.

To clarify whether the differences in the estimated stellar parameters Teff and log g severely affect the resulting abundance measurements, we repeated the [Fe/H], [Ca/Fe], and [Mg/Fe] determinations with FERRE, but after replacing the stellar parameters with those from SSPP. The resulting offset in the [Fe/H] determination compares better with the SSPP [Fe/H], although the dispersion increases slightly (δ = −0.09 dex, σ= 0.38 dex). Conversely, the dispersion in the comparison of the SSPP [α/Fe] with the new [Ca/Fe] and [Mg/Fe] estimates increases by more than 0.2 dex, although the offset for [Mg/Fe] is reduced from −0.14 dex to −0.01 dex. Thus, differences in Teff and log g are not solely responsible for the derived abundance contrast with the SSPP.

4.2. Comparison with high-resolution analyses

Several studies of EMP candidates from SDSS/SEGUE have been followed-up and analyzed with high-resolution spectra reported in the literature. Here we compare our present results with these measurements.

SDSS J031745.82+002304.1 was analyzed by Bonifacio et al. (2012) as one of their 16 EMP candidates found in the SDSS/SEGUE database. Our measurements for this star agree very well with their estimates from high-resolution spectra. We obtain Teff= 5780 K, log g = 3.72, [Fe/H] > −3.40, [Ca/Fe] ≥+ 0.62 and [Mg/Fe] ≥+ 0.21, while they obtained Teff= 5786 K, log g = 4.02, [Fe/H] = −3.46, [Ca/Fe] =+ 0.75 (+ 0.60 from Ca I lines) and [Mg/Fe] =+ 0.38.

Aoki et al. (2013) determined Teff, log g, [Fe/H], [Ca/Fe], and [Mg/Fe] for 70 VMP and EMP stars selected from SDSS/SEQUE. Nine of these objects are in common with our sample: [Fe/H] measurements from both analyses are available for seven stars, [Ca/Fe] for five stars, and [Mg/Fe] for six stars. Our Teff estimates are offset by −241 K from the Aoki et al. results, with a dispersion σ= 169 K. The log g comparison exhibits a larger offset and dispersion, δ = −1.03 dex and σ = 1.21 dex, similar to the comparison obtained with respect to the SSPP results. The [Fe/H] estimates compare reasonably well, with an offset of δ = −0.12 dex and σ = 0.21 dex. The [Mg/Fe] results are only in fair agreement, δ = −0.08 dex and σ = 0.30 dex. In contrast, our [Ca/Fe] estimates are 0.43 dex higher, with a dispersion of σ = 0.39 dex. However, the Aoki et al. determination of [Ca/H] in some of their stars came from the Ca I 4226 Å equivalent width, which is more sensitive to non-LTE effects than the Ca II HK doublet we considered.

We identify SDSS J132250.59+012342.9 as an EMP star, as previously reported by Placco et al. (2015). For this star we obtained the following parameters: Teff= 5234 K, log g= 0.84, [Fe/H] = −3.32, [Ca/Fe] = + 0.23 and [Mg/Fe] = + 0.30. Placco et al. obtained a lower Teff = 5008 ± 100 K, a higher log g = 1.95 ± 0.20, and a lower [Fe/H] = −3.64 ± 0.05, which are in reasonable agreement with our estimates. They also obtained [Ca/Fe] =+ 0.23 ± 0.08 and [Mg/Fe] = + 0.25 ± 0.05, which are also in good agreement with our results.

Recently, Susmitha Rani et al. (2016) reported [Fe/H] = −3.42 ± 0.19 for the star SDSS J134338.67+484426.6. This star is also included in our sample; we obtained estimates of Teff= 5307 K, log g= 0.51, [Fe/H] < −3.7, [Ca/H] = −3.40, and [Mg/H] = −3.52. Their temperature estimate is higher than ours, Teff= 5620 K, and their log g= 3.44 is considerably larger. We only determined an [Fe/H] upper limit, which is lower than their [Fe/H] measurement. They also determined [Ca/H] = −3.23 ± 0.16, which is consistent with our estimate, and [Mg/H] = −3.27 ± 0.16, slightly lower than our estimate.

In conclusion, the several comparisons we performed reveal that our temperature estimates are generally lower than previously reported results, by about 200−250 K, which is due to our use of spectroscopic estimates, a significant underestimate in our determination of log g, a slight underestimate of [Fe/H] (understandable from the Teff offset), and a dispersion ~0.3 dex in the derived [Ca/Fe] and [Mg/Fe].

5. Results

Our final sample of EMP stars is listed in Table 4, along with their ugriz magnitudes and heliocentric velocities, calculated by the SSPP (the typical accuracy for these velocities is on the order of 5 km s-1). Tables 5 and 6 list our estimated stellar atmospheric parameters, Teff,log g, and [M/H], and the derived abundances [Fe/H], [Ca/H], and [Mg/H], as well as their uncertainties. The minimum χ2 searches with FERRE were repeated ten times, with added random noise. The uncertainties are estimated from the standard deviation. We only list stars for which we obtained −4.0< [Fe/H] < −3.0. When our estimates were lower than the upper limit associated with the stellar parameters and S/N of the spectrum from which it was measured, a corresponding upper limit is stated. We obtained [Fe/H] measurements for 44 SDSS/SEGUE spectra and 4 BOSS spectra, 48% and 24% of each sample, respectively. For [Mg/H], we determined reliable estimates for 86 (93%) and 13 (76%) of the stellar spectra in the SDSS/SEGUE and BOSS samples, respectively. The [Ca/H] estimates were all reliable measurements. As noted above, Fe is the most difficult abundance to measure because its lines are so weak.

5.1. [Ca/Fe] and [Mg/Fe]

The α-element enrichment is of particular interest in this extremely low-metallicity regime, since it can provide valuable information on the nucleosynthesis histories of the first generations of stars. Figure 4 shows our derived [Ca/Fe] and [Mg/Fe] (split into SDSS/SEGUE and BOSS stars in the top and bottom panels) as a function of [Fe/H]. We use arrows to indicate upper limits on [Fe/H] (hence lower limits on [Ca(Mg)/Fe]). When neither Ca (or Mg) nor Fe abundances were determined, we plot their ratio with an asterisk. Overall, it appears that most of the stars present a value of [Ca(Mg)/Fe] consistent with the expected halo-star value of ~+0.4. From inspection of this figure, it is apparent that at the lowest metallicities, more stars lie above [α/Fe] =+ 0.4 than below this value.

From the stars for which we obtained reliable determinations of the abundances, we took into account results for which the deviation from [Ca/Fe] = + 0.4 was higher than three times the uncertainties of the ratios over the metallicity, or the iron abundance, in each case. We found two stars that exhibit ratios of [Ca/Fe] and [Mg/Fe] significantly higher than + 0.4 among the pre-BOSS stars: SDSS J035622.42+114705.4 ([Ca/Fe] =+ 0.68 ± 0.09; [Mg/Fe] =+ 0.87 ± 0.11), and SDSS J031259.10-061957.1 ([Ca/Fe] =+ 0.89 ± 0.10; [Mg/Fe] =+ 0.89 ± 0.14). These stars are shown in Fig. 4 as green dots. However, we note that the SSPP estimates for these stars are [α/Fe] =+ 0.14 and +0.41, respectively. Hence, more accurate chemical-abundance estimates are required to confirm their status as α-enhanced stars.

thumbnail Fig. 4

[Ca/Fe] and [Mg/Fe] ratios as a function of [Fe/H] from the analysis of our SDSS/SEGUE and BOSS samples. The reliable estimates after applying the upper limit evaluation are indicated as black dots. Red arrows show the cases where only an upper limit for the abundances could be obtained, and the direction the ratio would be situated in this diagram. The two stars for which we obtain high [Ca/Fe] and [Mg/Fe] are shown as green dots.

5.2. [C/Fe]

The SSPP provides [C/Fe] measurements for 57 stars in our sample with detected carbon. After examining the resulting values, we find 28 carbon-enhanced metal-poor (CEMP) stars ([C/Fe] > +0.7), from which we obtain a cumulative frequency of CEMP stars below [Fe/H] = −3.0 of ~26%. We also verified that the frequency of CEMP stars increases as the metallicity decreases. Dividing our sample at the median metallicity for the stars with measured metallicity ([Fe/H] = −3.2), we find that ~26% of the stars with −3.2≤ [Fe/H] ≤ −3.0 are CEMP stars, while ~39% of the stars below [Fe/H] = −3.2 are CEMP stars. We also note that below a metallicity of −3.2, the great majority of the stars shown in Fig. 5 are indeed CEMP stars, while the non-CEMP stars are the dominant fraction above this metallicity. Since more than half of our stars have only upper limits for [Fe/H], we repeated the exercise with the SSPP estimates of [Fe/H] for our full sample. For [Fe/H] ≤ −3.0, we obtain a cumulative frequency of CEMP stars of ~32%; for [Fe/H] ≤ −3.5, the cumulative frequency increases to ~42%. These results are consistent with the previous calculation within Poisson errors on the fractions (as a result of the small sample sizes involved, these are on the order of 10%).

There is no evidence of a correlation between [C/Fe] and [Ca(Mg)/Fe], nor with the SSPP [α/Fe] estimates, as Fig. 6 shows. Two stars exhibit high [C/Fe] (>+ 0.9) as well as [Mg/Fe] (>+ 0.6), but normal [Ca/Fe] and [α/Fe].

thumbnail Fig. 5

[C/Fe] estimates from the SSPP as a function of our [Fe/H] estimate. The red line indicates the level of [C/Fe] >+ 0.7 used to evaluate whether a star is considered carbon enhanced.

thumbnail Fig. 6

[C/Fe] estimates from the SSPP as a function of our [Ca/Fe] and [Mg/Fe] estimates (upper two panels, respectively). The lower panel shows [C/Fe] as a function of the SSPP [α/Fe] estimates. The red line indicates the level of [C/Fe] >+ 0.7 used to evaluate whether a star is considered carbon enhanced.

5.3. SDSS J134144.61+474128.6

We call attention to the CEMP star SDSS J134144.61+474128.6 in our sample, which is a bright (g = 11.90) star with [Fe/H] = −3.27 (the SSPP estimate of [Fe/H] = −2.95) and [C/Fe] =+ 0.95, identified during the course of the target search carried out for the MARVELS subsurvey (see the discussion of the MARVELS pre-survey in Rani et al. 2016). It is of interest that this star also exhibits elevated magnesium ([Mg/Fe] =+ 0.62), which is often found for CEMP-no stars (see, e.g., Norris et al. 2013). The absolute carbon abundance, A(C) = log  ϵ = 6.11, places it on the low-C band that is associated with most CEMP-no stars (see, e.g., Bonifacio et al. 2015; Yoon et al., in prep.). We note that the other bright EMP star in our sample, discussed by Rani et al. (2016; SDSS J134338.67+484426.6), is not carbon enhanced. Only ~20 EMP stars have been identified to date that are as bright as these two stars, hence we plan to obtain a spectrum with higher resolution and higher S/N of SDSS J1341+4741 in the near future.

6. Conclusions

We presented stellar atmospheric parameters and abundance estimates of [Fe/H], [Mg/H], and [Ca/H] for 108 extremely metal-poor stars from SDSS/SEGUE and BOSS with iron abundances in the range −4.0< [Fe/H] < −3.0. Below we summarize our main conclusions:

  1. The determination of Fe and Mg abundances for EMP stars from individual lines in spectra at low spectral resolution (R ~ 2000) depends critically on the signal-to-noise ratio because these lines are weak in this metallicity range. Conversely, the Ca II HK resonance lines in the blue region of the spectrum are sufficiently strong to reliably quantify the abundance of Ca. We established a relation between S/N and [X/H] to determine upper limits on the abundances of Fe, Ca, and Mg when a reliable estimate was not possible.

  2. We derived an analytical function that reproduced the detection limits for different elements as a function of S/N. These curves can be well fit by a parabola and a linear polynomial, with coefficients that depend on Teff and log g. We reported analytical functions that specify the lowest required S/N to reliably estimate [Fe/H], [Ca/H], and [Mg/H] in spectra with R ~ 2000 and S/N< 90 from the spectral regions 4885 <λ< 5070 Å, 5220 <λ< 5280 Å, and 5295 <λ< 5500 Å (to determine Fe); 5160 <λ< 5190 Å (to determine Mg abundances); and 3910 <λ< 3990 Å (for Ca abundances).

  3. We determined [Ca/Fe] and [Mg/Fe] abundance ratios for our program stars. The overall trend with metallicity is consistent with an α-element enhancement [α/Fe] ~ +0.4 dex. However, several stars for which only Fe upper limits were estimated indicate high [α/Fe] ratios, mainly at the lowest metallicities considered.

  4. The [C/Fe] estimates obtained with the SSPP revealed a cumulative frequency of ~26% CEMP stars for [Fe/H] < −3.0, comparable to that found by Lee et al. (2013), 28%. The frequency of CEMP stars also increases with decreasing metallicity, as reported by previous studies. We found no evidence for a [C/Fe] correlation with [Ca/Fe], [Mg/Fe], nor the SSPP [α/Fe] measurements.

  5. We identified a bright (g = 11.90) EMP star in our sample, SDSS J134144.61+474128.6, with enhanced [C/Fe] and [Mg/Fe], as well as a low absolute carbon abundance, A(C)=6.11, a pattern typically associated with CEMP-no stars. Higher resolution spectroscopic follow-up of this star is planned.

Several stars have reported [α/Fe] significantly higher than + 0.4 in the literature. For example, Aoki et al. (2007) reported on the highly α-element enhanced VMP star from the HK survey, BS 16934-002, with [Mg/Fe] =+ 1.23, but normal [Ca/Fe] = +0.44, and no carbon overabundance. Other works reported high [C/Fe] and [Mg/Fe], but little evidence of [Ca/Fe] enhancement with respect to the typical halo values (Norris et al. 2013; Yong et al. 2013; Hansen et al. 2015). Two scenarios were proposed for this chemical pattern: i) the results of mixing and processing of material due to stellar rotation; or ii) nucleosynthesis in mixing and fallback supernova models. Elements such as Ca and Si provide the key to understanding which of these possibilities is more likely. We only found two stars for which both [C/Fe] and [Mg/Fe] are enhanced but that exhibit normal [Ca/Fe], the expected chemical pattern for a massive spinstar (Norris et al. 2013; Maeder et al. 2015).

Puzia et al. (2006) measured [α/Fe] ratios significantly higher than +0.5 for globular clusters in early-type elliptical galaxies based on Lick line-index measurements, at −1< [Z/H] <0 – these authors used [Z/H] to indicate the global metallicity in a galaxy, estimated from Mg and Fe lines (see Puzia et al. 2006; González 1993). They suggested that massive stars are the potential progenitors, with M>20 M⊙, or with M~ 130−190 M⊙ that explode as pair-instability SNe. Both possibilities imply extremely short timescales, on the order of few Myr. Therefore, the authors concluded that these stars may belong to the first generation of star clusters formed in their respective galaxies. In Paper II we reported high [Ca/Fe] and [Mg/Fe] median values for stars in the outer-halo region of the Galaxy, at Galactocentric radii greater than 40 kpc. Such stars could have been formed in low-mass fragments at very early stages of the evolution of the Milky Way, and later accreted into the Galactic halo (see, e.g., Tissera et al. 2014, and references therein).

In contrast, Caffau et al. (2013a) found three stars with [Fe/H] < −3.0 with low [α/Fe] ratios. Similarly low [α/Fe] stars had previously been detected by Nissen & Schuster (2010). However, the latter authors found these ratios for stars with higher [Fe/H], and interpreted this population as being born after the explosion of Type Ia SNe. In this scenario, the low [α/Fe] ratios would then be the result of the addition of Fe from low- to intermediate-mass stars. However, at metallicities lower than [Fe/H] = −3.0, few SNIa explosions are expected to have occurred. The interpretation offered by Caffau et al. (2013a,b) is that two starbursts could have taken place in their progenitor fragment; the [low-α/Fe] ratios could then have resulted from gas enriched by SNIa explosions of stars formed in the first burst. This hypothesis was also invoked by Carigi et al. (2002) to explain low [O/Fe] in Milky Way dwarf spheroidal satellites from chemical evolution models.

It would be desirable to obtain more accurate estimates to refine our reported α-element enhancements in cases where the iron abundance could not be determined. The analysis performed in Paper II revealed high [α/Fe] enhancements for very metal-poor stars in the outer-halo region of our Galaxy. Accurate [Fe/H] estimates for our stars would allow us to test whether these results are confirmed by our new sample.

Acknowledgments

E.F.A. acknowledges support from DGAPA-UNAM postdoctoral fellowships. C.A.P. acknowledges support from the Spanish MINECO through grant AYA2014-56359-P. T.C.B. acknowledges partial support for this work from grants PHY 08-22648; Physics Frontier Center/Joint Institute or Nuclear Astrophysics (JINA), and PHY 14-30152; Physics Frontier Center/JINA Center for the Evolution of the Elements (JINA-CEE), awarded by the US National Science Foundation. Y.S.L. acknowledges partial support from the National Research Foundation of Korea to the Center for Galaxy Evolution Research and Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT & Future Planning (NRF-2015R1C1A1A02036658). Funding for SDSS-III has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, and the U.S. Department of Energy Office of Science. The SDSS-III web site is http://www.sdss3.org/. SDSS-III is managed by the Astrophysical Research Consortium for the Participating Institutions of the SDSS-III Collaboration including the University of Arizona, the Brazilian Participation Group, Brookhaven National Laboratory, University of Cambridge, Carnegie Mellon University, University of Florida, the French Participation Group, the German Participation Group, Harvard University, the Instituto de Astrofisica de Canarias, the Michigan State/Notre Dame/JINA Participation Group, Johns Hopkins University, Lawrence Berkeley National Laboratory, Max Planck Institute for Astrophysics, Max Planck Institute for Extraterrestrial Physics, New Mexico State University, New York University, Ohio State University, Pennsylvania State University, University of Portsmouth, Princeton University, the Spanish Participation Group, University of Tokyo, University of Utah, Vanderbilt University, University of Virginia, University of Washington, and Yale University.

References

  1. Alam, S., Albareti, F. D., Allen de Prieto, C., et al. 2015, ApJS, 219, 12 [NASA ADS] [CrossRef] [Google Scholar]
  2. Allen de Prieto, C., Beers, T. C., Wilhelm, R., et al. 2006, ApJ, 636, 804 [NASA ADS] [CrossRef] [Google Scholar]
  3. Allen de Prieto, C., Sivarani, T., Beers, T. C., et al. 2008, AJ, 136, 2070 [NASA ADS] [CrossRef] [Google Scholar]
  4. Allen de Prieto, C., Fernández-Alvar, E., Schlesinger, K. J., et al. 2014, A&A, 568, A7 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  5. Allen de Prieto, C., Fernández-Alvar, E., Aguado, D. S., et al. 2015, A&A, 579, A98 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  6. An, D., Beers, T. C., Johnson, J. A., et al. 2013, ApJ, 763, 65 [NASA ADS] [CrossRef] [Google Scholar]
  7. An, D., Beers, T. C., Santucci, R. M., et al. 2015, ApJ, 813, L28 [NASA ADS] [CrossRef] [Google Scholar]
  8. Aoki, W., Honda, S., Beers, T. C., et al. 2005, ApJ, 632, 611 [NASA ADS] [CrossRef] [Google Scholar]
  9. Aoki, W., Honda, S., Beers, T. C. et al. 2007, ApJ, 660, 747 [NASA ADS] [CrossRef] [Google Scholar]
  10. Aoki, W., Beers, T. C., Lee, Y. S., et al. 2013, AJ, 145, 13 [NASA ADS] [CrossRef] [Google Scholar]
  11. Arnone, E., Ryan, S. G., Argast, et al. 2005, A&A, 430, 507 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  12. Auer, L. 2003, Stellar Atmosphere Modeling, 288, 3 [NASA ADS] [Google Scholar]
  13. Barklem, P. S., Christlieb, N., Beers, T. C., et al. 2005, A&A, 439, 129 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  14. Beers, T. C., Preston, G. W., & Shectman, S. A. 1985, AJ, 90, 2089 [NASA ADS] [CrossRef] [Google Scholar]
  15. Beers, T. C., Preston, G. W., & Shectman, S. A. 1992, AJ, 103, 1987 [NASA ADS] [CrossRef] [Google Scholar]
  16. Beers, T. C., Carollo, D., Ivezić, Ž., et al. 2012, ApJ, 746, 34 [NASA ADS] [CrossRef] [Google Scholar]
  17. Bonifacio, P., Spite, M., Cayrel, R., et al. 2009, A&A, 501, 519 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  18. Bonifacio, P., Sbordone, L., Caffau, E., et al. 2012, A&A, 542, A87 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  19. Bonifacio, P., Caffau, E., Spite, M., et al. 2015, A&A, 579, A28 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  20. Caffau, E., Bonifacio, P., François, P., et al. 2011a, Nature, 477, 67 [NASA ADS] [CrossRef] [PubMed] [Google Scholar]
  21. Caffau, E., Bonifacio, P., François, P., et al. 2011b, A&A, 534, A4 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  22. Caffau, E., Bonifacio, P., François, P., et al. 2012, A&A, 542, A51 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  23. Caffau, E., Bonifacio, P., François, P., et al. 2013a, A&A, 560, A15 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  24. Caffau, E., Bonifacio, P., Sbordone, L., et al. 2013b, A&A, 560, A71 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  25. Carigi, L., Hernandez, X., & Gilmore, G. 2002, MNRAS, 334, 117 [NASA ADS] [CrossRef] [Google Scholar]
  26. Carollo, D., Beers, T. C., Lee, Y. S., et al. 2007, Nature, 450, 1020 [NASA ADS] [CrossRef] [PubMed] [Google Scholar]
  27. Carollo, D., Beers, T. C., Chiba, M., et al. 2010, ApJ, 712, 692 [NASA ADS] [CrossRef] [Google Scholar]
  28. Cayrel, R., Depagne, E., Spite, M., et al. 2004, A&A, 416, 1117 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  29. Chen, Y. Q., Zhao, G., Carrell, K., et al. 2014, ApJ, 795, 521 [Google Scholar]
  30. Christlieb, N. 2003, Rev. Mod. Phys., 16, 191 [Google Scholar]
  31. Cohen, J. G., Christlieb, N., McWilliam, A., et al. 2004, ApJ, 612, 1107 [NASA ADS] [CrossRef] [Google Scholar]
  32. Cohen, J. G., McWilliam, A., Christlieb, N., et al. 2007, ApJ, 659, L161 [NASA ADS] [CrossRef] [Google Scholar]
  33. Cohen, J. G., Christlieb, N., McWilliam, A., et al. 2008, ApJ, 672, 320 [NASA ADS] [CrossRef] [Google Scholar]
  34. Dawson, K. S., Schlegel, D. J., Ahn, C. P., et al. 2013, AJ, 145, 10 [Google Scholar]
  35. de Jong, J. T. A., Yanny, B., Rix, H.-W., et al. 2010, ApJ, 714, 663 [NASA ADS] [CrossRef] [Google Scholar]
  36. Eisenstein, D. J., Weinberg, D. H., Agol, E., et al. 2011, AJ, 142, 72 [Google Scholar]
  37. Fernández-Alvar, E., Allen de Prieto, C., Schlesinger, K. J., et al. 2015, A&A, 577, A81 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  38. Frebel, A., Aoki, W., Christlieb, N., et al. 2005, Nature, 434, 871 [NASA ADS] [CrossRef] [PubMed] [Google Scholar]
  39. Frebel, A., Chiti, A., Ji, A. P., Jacobson, H. R., & Placco, V. M. 2015, ApJ, 810, L27 [NASA ADS] [CrossRef] [Google Scholar]
  40. González, J. J. 1993, Ph.D. Thesis, University of California, Santa Cruz [Google Scholar]
  41. Gunn, J. E., Siegmund, W. A., Mannery, E. J., et al. 2006, AJ, 131, 2332 [NASA ADS] [CrossRef] [Google Scholar]
  42. Hansen, T., Hansen, C. J., Christlieb, N., et al. 2015, ApJ, 807, 173 [NASA ADS] [CrossRef] [Google Scholar]
  43. Janesh, W., Morrison, H. L., Ma, Z., et al. 2016, ApJ, 816, 80 [NASA ADS] [CrossRef] [Google Scholar]
  44. Keller, S. C., Bessell, M. S., Frebel, A., et al. 2014, Nature, 506, 463 [NASA ADS] [CrossRef] [PubMed] [Google Scholar]
  45. Lai, D. K., Rockosi, C. M., Bolte, M., et al. 2009, ApJ, 697, L63 [NASA ADS] [CrossRef] [Google Scholar]
  46. Lee, Y. S., Beers, T. C., Sivarani, T., et al. 2008a, AJ, 136, 2022 [NASA ADS] [CrossRef] [Google Scholar]
  47. Lee, Y. S., Beers, T. C., Sivarani, T., et al. 2008b, AJ, 136, 2050 [NASA ADS] [CrossRef] [Google Scholar]
  48. Lee, Y. S., Beers, T. C., Allen de Prieto, C., et al. 2011, AJ, 141, 90 [NASA ADS] [CrossRef] [Google Scholar]
  49. Lee, Y. S., Beers, T. C., Masseron, T., et al. 2013, AJ, 146, 132 [NASA ADS] [CrossRef] [Google Scholar]
  50. Li, H., Aoki, W., Zhao, G., et al. 2015, PASJ, 67, 84 [NASA ADS] [CrossRef] [Google Scholar]
  51. Maeder, A., Meynet, G., & Chiappini, C. 2015, A&A, 576, A56 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  52. Meléndez, J., Placco, V. M., Tucci-Maia, M., et al. 2016, A&A, 585, L5 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  53. Nissen, P. E., & Schuster, W. J. 2010, A&A, 511, L10 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  54. Nissen, P. E., & Schuster, W. J. 2011, A&A, 530, A15 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  55. Norris, J. E., Bessell, M. S., Yong, D., et al. 2013, ApJ, 762, 25 [NASA ADS] [CrossRef] [Google Scholar]
  56. Placco, V. M., Frebel, A., Lee, Y. S., et al. 2015, ApJ, 809, 136 [NASA ADS] [CrossRef] [Google Scholar]
  57. Puzia, T. H., Kissler-Patig, M., & Goudfrooij, P. 2006, ApJ, 648, 383 [NASA ADS] [CrossRef] [Google Scholar]
  58. Reimers, D., & Wisotzki, L. 1997, The Messenger, 88, 14 [NASA ADS] [Google Scholar]
  59. Roederer, I. U. 2009, AJ, 137, 272 [NASA ADS] [CrossRef] [Google Scholar]
  60. Roederer, I. U., Schatz, H., Lawler, J. E., et al. 2014, ApJ, 791, 32 [NASA ADS] [CrossRef] [Google Scholar]
  61. Smee, S. A., Gunn, J. E., Uomoto, A., et al. 2013, AJ, 146, 32 [Google Scholar]
  62. Smolinski, J. P., Lee, Y. S., Beers, T. C., et al. 2011, AJ, 141, 89 [NASA ADS] [CrossRef] [Google Scholar]
  63. Susmitha Rani, A., Sivarani, T., Beers, T. C., et al. 2016, MNRAS, 458, 2648 [NASA ADS] [CrossRef] [Google Scholar]
  64. Tissera, P., Beers, T. C., Carollo, D., et al. 2014, MNRAS, 439, 2128 [Google Scholar]
  65. Yanny, B., Rockosi, C., Newberg, H. J., et al. 2009, AJ, 137, 4377 [NASA ADS] [CrossRef] [Google Scholar]
  66. Yong, D., Norris, J. E., Bessell, M. S., et al. 2013, ApJ, 762, 27 [CrossRef] [Google Scholar]
  67. York, D. G., Adelman, J., Anderson, J. E., Jr., et al. 2000, AJ, 120, 1579 [Google Scholar]

All Tables

Table 1

Second-order polynomial coefficients that reproduce the parabolic and linear polynomial parameters as a function of Teff for each of the log g values considered, 1.5 and 4.2, to estimate Fe.

In the text
Table 2

Same as in Table 1, but for Mg.

In the text
Table 3

Same as in Table 1, but for Ca.

In the text

All Figures

thumbnail Fig. 1

Model fits for a SDSS/SEGUE star, SDSS J093339.24+310245.4, with the mean S/N of the sample, ~40. The three panels in the left column correspond to the fit regions from which the [Fe/H] abundance was determined. The derived iron abundance is shown in the legend of the upper panel, while the temperature and surface gravity estimates are shown in the legend of the lower panel. The panels in the right column show the CaII HK doublet and the MgIb triplet that were fit to determine Ca and Mg abundances, with the derived estimates shown in the legends.
In the text
thumbnail Fig. 2

χ2 obtained from the evaluation of a simulated spectrum at [M/H]0= [Fe/H]0= −3.8 (marked with a vertical dotted line) compared with synthetic spectra over a range of metallicities [M/H] (from −5.0 to −2.5 with a step of 0.1 dex). The blue line shows the resulting χ2 curve for a simulation with S/N = 30, and the red line applies to S/N = 80. The two black dotted lines indicate the linear fits up to the [M/H]0 corresponding to the minimum χ2, from which the slope and its uncertainty are derived.
In the text
thumbnail Fig. 3

Lowest required signal-to-noise values, S/N0, to reliably determine [Fe/H]0, [Mg/H]0 and [Ca/H]0 abundances, evaluated at log g = 1.5 and log g = 4.5 and 4750 < Teff < 6500 K (darker colors correspond to lower temperatures). The increase in Teff complicates the [Fe/H] determination, mainly for Fe and Mg, and a higher S/N is necessary. The sensitivity of the Ca II HK resonance lines enables determining [Ca/H] at very low abundances.
In the text
thumbnail Fig. 4

[Ca/Fe] and [Mg/Fe] ratios as a function of [Fe/H] from the analysis of our SDSS/SEGUE and BOSS samples. The reliable estimates after applying the upper limit evaluation are indicated as black dots. Red arrows show the cases where only an upper limit for the abundances could be obtained, and the direction the ratio would be situated in this diagram. The two stars for which we obtain high [Ca/Fe] and [Mg/Fe] are shown as green dots.
In the text
thumbnail Fig. 5

[C/Fe] estimates from the SSPP as a function of our [Fe/H] estimate. The red line indicates the level of [C/Fe] >+ 0.7 used to evaluate whether a star is considered carbon enhanced.
In the text
thumbnail Fig. 6

[C/Fe] estimates from the SSPP as a function of our [Ca/Fe] and [Mg/Fe] estimates (upper two panels, respectively). The lower panel shows [C/Fe] as a function of the SSPP [α/Fe] estimates. The red line indicates the level of [C/Fe] >+ 0.7 used to evaluate whether a star is considered carbon enhanced.
In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.