A Search for Strongly Mg-enhanced Stars from the Sloan Digital Sky Survey
RResearch in Astronomy and Astrophysics manuscript no.(L A TEX: lx-raav2.tex; printed on May 20, 2014; 15:14)
A Search for Strongly Mg-enhanced Stars from the Sloan Digital SkySurvey
Xiang Li , , Gang Zhao , Yu-Qin Chen , Hai-Ning Li Key Laboratory of Optical Astronomy, National Astronomical Observatories, Chinese Academy ofSciences, Beijing 100012, China; [email protected] University of Chinese Academy of Sciences, Beijing 100049, China
Abstract
Strongly Mg-enhanced stars with [Mg/Fe] > ( R ≃ spectra of the Sloan Digital Sky Survey (SDSS) is carried out byfinding the best matched synthetic spectrum to the observed one in the region of Mg I b linesaround λ > Key words: stars: abundance — stars: chemically peculiar — method: data analysis —techniques: spectroscopic analysis
Enhancements of α -elements in metal-poor stars have been first identified by Aller & Greenstein (1960),and later Wallerstein (1962) has confirmed this enhancement as a typical [ α /Fe] ratio of ∼ + ∼− X. Li et al. that metal-poor halo stars almost have a constant [ α /Fe] ratio of ∼ + α /Fe] ratios ofthick-disk stars( ∼ + + ∼ + + + α -elements including Ca and Ti, aswell as light odd-Z elements such as Na and Al, but exhibit peculiar abundance of neutron-capture elements.It seems that the nucleosynthesis mechanisms for these strongly Mg-enhanced stars are quite different fromthose of normal stars.It has been suggested that there are three categories of strongly Mg-enhanced stars. (1) Carbon-enhanced metal-poor stars with large overabundant elements produced by the slow- and rapid-neutron-capture process (CEMP-rs). Mg and Mg are produced significantly in high-mass AGB stars duringa convective s-process driven by the Ne( α ,n) Mg neutron source (Goriely & Siess 2005; Karakas &Lattanzio 2003), which is responsible for the observed high Mg (= Mg+ Mg+ Mg) abundance in someCEMP-rs stars (Masseron et al. 2010). (2)
Carbon-enhanced metal-poor stars with no enhancement ofneutron-capture elements (CEMP-no). CEMP-no stars are born out of gas with large amount of C, which ispolluted by low-energy faint supernova, and they have undergone the first dredge-up and processed certainamount of pristine C into N (Ryan et al. 2005). The high [Mg/Fe] ratio of CEMP-no stars can be explainedwith the “mixing and fallback” model by Umeda & Nomoto (2005), which suggests that the high [Mg/Fe]ratio occurs if the mixing-fallback region (with large amount of ejected Fe) does not extend beyond the Mglayers (Tsujimoto & Shigeyama 2003). (3) α -enhanced metal-poor star with no enhancement of carbon andneutron-capture elements (AEMP). The explanation for the only AEMP star, BS 16934-002, is analogous tothat of CEMP-no stars, except that the non-enhancement of C is due to the effect of significant mass loss ofout layers containing C-rich material produced from its massive progenitor (Aoki et al. 2007b). The originsof the strongly Mg-enhanced stars and their abundance patterns provide us crucial evidence to understandthe early chemical evolution of the Galaxy.However, the number of strongly Mg-enhanced stars identified by now is still very limited and thereis not yet any systematic search for such stars. The large database from the SDSS spectroscopic survey(York et al. 2000) provides us an unprecedented opportunity to conduct such a systematic investigation ofstrongly Mg-enhanced stars; therefore our work aims to utilize the spectroscopic data from SDSS DR9 toconduct a systematic search for strongly Mg-enhanced stars (Ahn et al. 2012). This paper is organized asfollows. In section 2, we give a brief description on the sample selection. Section 3 describes the methodfor [Mg/Fe] determinations. A list of the candidates of strongly Mg-enhanced stars is presented in Section4, and Section 5 is a summary of the main results. The spectroscopic data and atmospheric parameters we have used are based on SDSS DR9. Although thelatest data release of DR10 (Ahn et al. 2014) is available now, there is no update on the low-resolution stellar
Search for Strongly Mg-enhanced Stars from Sloan Digital Sky Survey 3 spectroscopy and thus it would make no difference if the sample is re-selected from DR10. The selectionprocedure is as follows. Firstly, F and G-type stars with . < ( g − r ) < . (Yanny et al. 2009) andS/N at g-band larger than 30 are selected. Secondly, based on the stellar parameters determined by SSPP(SEGUE Stellar Parameter Pipeline, Lee et al. 2008a,b; Allende Prieto et al. 2008), further selection aremade to include only stars with [Fe/H] < − dex and 5000 K < T eff < < g − r < ∼
50 K for T eff , ∼ dex for log g , and ∼ dex for [Fe/H], respectively. These errors increase to ∼
80 K, ∼ dex , and ∼ dex for T eff , log g and [Fe/H] for stars with − < g − r < < − <
15. Finally, we estimate the S/N around Mgb lines, S/N Mg from SDSS spectra inthe wavelength range of 5280 − >
30 (given by SSPP) are not withsufficient spectral quality at Mg I b region which is important for our work. Thus, we exclude stars withS/N Mg <
30. The above selection results in a final sample of 14,850 stars.
Spectral synthesis of the Mg I b feature is carried out to derive the [Mg/Fe] ratios of the sample with thelocal thermodynamic equilibrium (LTE) atmospheric models. Although Lee et al. (2011) have determined[ α /Fe] ratios from the SDSS spectra of these stars, values of individual stars are not publically availableat present. Moreover, their [ α /Fe] ratios are derived from four α -elements, Mg, Ti, Si and Ca over thewide wavelength range from 4500-5500 ˚A by using the weighting factors 5, 3, 1 and 1, respectively.The general matching to a wide wavelength range of spectra may be a good approximation for estimating[ α /Fe], but has no advantage in searching for strongly Mg-enhanced candidates since the contributionsfrom other elements in the wide wavelength range will possibly counterbalance the contribution of theenhancement of Mg elements (if exists) to an undetected level. This is also pointed out by Lee et al. (2011)that such measurement may not correctly represent the overall content of the α -elements, especially in casesof abnormally high or low Mg abundances.In this paper, we aim to derive [Mg/Fe] ratios from the narrow wavelength range around the Mg I bfeature, which is dominated by Mg element, with a specific purpose of picking out strongly Mg-enhancedcandidates. A line-profile-matching method is performed on the three Mg lines individually by varying[Mg/Fe] ratios of the synthetic spectra from − + χ values of a set of [Mg/Fe] ratios are fittedwith a third order polynomial fit and the [Mg/Fe] corresponding with the minimum χ value is consideredto be the best-fit value. The matching procedure with χ method is widely used, but the choice of differentfitting parameters is key to the derived abundances. In the following sections, we will focus on our effortsin refining these fitting parameters by using the spectra and the results from high-resolution analysis. In order to set the fitting parameters and to check whether the Mg I b feature is a valid and robust proxy forestimating [Mg/Fe] ratio, we apply our matching method to a reference sample of 47 dwarf stars (Nissen &
X. Li et al.
Schuster 2010), (hereafter NS10), with high-resolution and high-S/N FIES (FIbre fed Echelle Spectrograph)spectra (kindly provided by Poul Nissen) together with accurate atmospheric parameters and very highprecision Mg abundances. The FIES spectra of 47 reference stars cover a wavelength range from 4000 ˚A to7000 ˚A with a resolution of R ≃ S/N ≃ − . The code adopts the one-dimensional, 72-layer, plane-parallel and line-blanketed modelswithout convective overshoot, which are linearly interpolated over α -enhanced AODFNEW grid (Castelli& Kurucz 2003). The [ α /Fe] is assumed to be . dex when [Fe/H] < − . dex in these models.The detailed procedure of our method is as follows. According to atmospheric parameters provided byNS10, corresponding atmosphere model is generated by linear interpolations with Castelli and Kurucz’smodel grid. After that, a set of synthetic spectra are produced by SPECTRUM for various [Mg/Fe] ratiosfrom − + R ≃ χ values are calculated from the deviation between the synthetic and the observed spectra. A third order poly-nomial is used to fit χ values, and the [Mg/Fe] ratio corresponding to the minimum χ value is consideredto be the best-fit value. Before applying our method to the SDSS spectra, we would like to refine and testify our method with high-resolution and high-S/N FIES spectra. Firstly, the line list and atomic data of the Mg I b region shouldbe checked. An isotope-compatible line list provided by SPECTRUM is adopted as it is updated recently(private communication with Richard O. Gray). However, we find that the log gf value of Fe 5162.292 ˚Ais not suitable, which will blend the first line of Mg I b in low-resolution spectra. Because the syntheticspectra with accurate Fe abundances from NS10 do not match this iron line for any of the 47 FIES spectra.We thus adopt the log gf value of this line from Lambert et al. (1996), which results in a good agreementbetween the observed and synthetic spectra. Secondly, we check the reliability of [Mg/Fe] derived fromMg I b compared with other two magnesium lines, Mg 4730.04 ˚A and Mg 5711.10 ˚A which are oftenused in abundance analysis of high-resolution spectra. We apply our line-profile-matching procedure to Mg4730.04 ˚A , Mg 5711.10 ˚A and Mg I b lines of FIES spectra with R ≃ . dex . The [Mg/Fe] derived from these threelines are presented in columns 7 to 9 of Table 1, and the deviation between the [Mg/Fe] determined fromMg I b and the other two magnesium lines is about 0.05 dex . According to Gehren et al. (2006), suchdeviation can be explained by the NLTE effect which is around 0.065 dex . Thirdly, we apply our method tothe smoothed FIES spectra with R ≃ R ≃ R ≃ α abundances based onhigh-resolution spectra observed with the Hobby Eberly Telescopy (Lee et al. 2011). We adjust the smooth Search for Strongly Mg-enhanced Stars from Sloan Digital Sky Survey 5
Fig. 1
Comparison of a SDSS spectrum and corresponding synthetic spectra with different s-mooth factors (solid line for 2.2 and dash line for 2.8)factor until the smoothed synthetic spectrum well matches the observed one and a final value of 2.2 isadopted as the optimum value as shown in Figure 1. The [Mg/Fe] derived from the smoothed FIES spectrawith R ≃ ± − and derive [Mg/Fe] ratios with thesmoothed FIES spectra. It is found that such variations in the micro-turbulence will not significantly affectthe [Mg/Fe] derived from the smoothed FIES spectra of NS10. Figure 3 presents the difference of [Mg/Fe]rations derived from the low-resolution spectra due to a change of ± − in micro-turbulence for the47 FIES stars in NS10. The resulted deviation in [Mg/Fe] is less than 0.03 dex . Some Galactic globular cluster with precise abundance analysis of high-resolution have been observed bySDSS. Therefore, it will be a good test to apply our method on the SDSS spectra of member stars ofglobular cluster. Member stars from four Galactic globular clusters (M92, M13, M3, M71) with a coveragein the metallicity space comparable to our sample are carefully chosen to test the validation of our method.Member stars in clusters are supposed to be born simultaneously and exhibit similar elemental-abundancepatterns due to the well-mixed interstellar medium at the same location in the Galaxy. Although the anti-correlations of Na-O and Mg-Al in globular clusters (GCs) indicate significant star-to-star variations ofthese light elements, according to Carretta et al. (2009), the variation of Mg abundances in 19 normal GCsis around 0.07 dex for HB stars. The four GCs adopted here have been well investigated in the literatureand the scatters in their [Mg/Fe] ratios are around 0.1 − dex as shown in Table 2, which may be mainly X. Li et al.
Table 1
Comparison of [Mg/Fe] for different Mg lines and dif-ferent resolutions of Mg I b feature for stars in NS10
ID T eff log g [Fe/H] Turb [Mg/Fe] NS HRS-4703 HRS-5711 HRS-Mg I b LRS-Mg I bG05-36 6013 4.23 -1.233 1.39 0.33 0.34 0.33 0.24 0.24G119-64 6181 4.18 -1.477 1.50 0.25 0.23 0.20 0.10 0.12G125-13 5848 4.28 -1.434 1.50 0.30 0.30 0.20 0.20 0.21G127-26 5791 4.14 -0.529 1.21 0.31 0.21 0.34 0.20 0.19G13-38 5263 4.54 -0.876 0.92 0.36 0.34 0.45 0.35 0.35G15-23 5297 4.57 -1.097 1.00 0.40 0.34 0.44 0.39 0.39G150-40 5968 4.09 -0.807 1.41 0.14 0.18 0.18 0.09 0.08G16-20 5625 3.64 -1.416 1.51 0.22 0.25 0.25 0.20 0.21G161-73 5986 4.00 -0.999 1.36 0.13 0.04 0.10 0.00 0.00G170-56 5994 4.12 -0.924 1.49 0.18 0.14 0.17 0.05 0.06G172-61 5225 4.47 -1.000 0.86 0.16 0.22 0.24 0.25 0.25G176-53 5523 4.48 -1.337 1.00 0.15 0.21 0.21 0.18 0.19G180-24 6004 4.21 -1.393 1.55 0.31 0.29 0.29 0.23 0.24G187-18 5607 4.39 -0.666 1.15 0.29 0.27 0.35 0.26 0.26G192-43 6170 4.29 -1.339 1.45 0.19 0.20 0.17 0.09 0.09G20-15 6027 4.32 -1.485 1.60 0.22 0.21 0.27 0.16 0.15G21-22 5901 4.24 -1.089 1.40 0.09 0.06 0.06 0.02 0.01G232-18 5559 4.48 -0.928 1.26 0.36 0.34 0.40 0.31 0.31G24-13 5673 4.31 -0.721 0.96 0.34 0.24 0.40 0.26 0.26G24-25 5825 3.85 -1.402 1.13 0.35 0.37 0.39 0.21 0.21G31-55 5638 4.30 -1.097 1.36 0.28 0.34 0.37 0.32 0.31G49-19 5772 4.25 -0.552 1.22 0.30 0.24 0.35 0.23 0.22G53-41 5859 4.27 -1.198 1.30 0.24 0.28 0.22 0.15 0.14G56-30 5830 4.26 -0.891 1.32 0.09 0.07 0.14 0.05 0.04G56-36 5933 4.28 -0.938 1.43 0.20 0.17 0.24 0.12 0.13G57-07 5676 4.25 -0.474 1.09 0.34 0.25 0.40 0.27 0.26G74-32 5772 4.36 -0.724 1.14 0.37 0.27 0.39 0.26 0.25G75-31 6010 4.02 -1.035 1.38 0.12 0.14 0.10 0.05 0.04G81-02 5859 4.19 -0.689 1.31 0.25 0.15 0.26 0.11 0.11G85-13 5628 4.38 -0.586 0.97 0.33 0.27 0.38 0.26 0.25G87-13 6085 4.13 -1.088 1.52 0.10 0.16 0.08 0.05 0.06G94-49 5373 4.50 -0.796 1.10 0.35 0.33 0.39 0.36 0.35G96-20 6293 4.41 -0.889 1.52 0.30 0.26 0.24 0.17 0.17G98-53 5848 4.23 -0.874 1.30 0.20 0.19 0.24 0.17 0.15G99-21 5487 4.39 -0.668 0.89 0.33 0.28 0.42 0.28 0.27HD148816 5823 4.13 -0.731 1.43 0.32 0.27 0.37 0.23 0.23HD159482 5737 4.31 -0.726 1.31 0.34 0.29 0.38 0.26 0.25HD160693 5714 4.27 -0.487 1.12 0.24 0.20 0.34 0.18 0.17HD177095 5349 4.39 -0.737 0.85 0.38 0.27 0.46 0.30 0.30HD179626 5850 4.13 -1.041 1.57 0.35 0.34 0.41 0.25 0.27HD189558 5617 3.80 -1.121 1.39 0.36 0.32 0.40 0.27 0.27HD193901 5650 4.36 -1.090 1.22 0.13 0.13 0.18 0.11 0.11HD194598 5942 4.33 -1.093 1.40 0.18 0.16 0.22 0.09 0.10HD230409 5318 4.54 -0.849 1.11 0.30 0.29 0.38 0.33 0.33HD233511 6006 4.23 -1.547 1.30 0.36 0.32 0.27 0.29 0.29HD237822 5603 4.33 -0.450 1.09 0.35 0.27 0.41 0.28 0.27HD250792 5489 4.47 -1.013 1.08 0.23 0.20 0.25 0.20 0.24 Search for Strongly Mg-enhanced Stars from Sloan Digital Sky Survey 7
Fig. 2
Upper and lower left panels: The matching of Mg 4703 and Mg 5711 lines between FIESspectrum (solid lines) and the best matched synthetic spectrum (dash lines) for G176-53. Upperright panels: The matching of Mg I b lines between the FIES spectrum (solid lines) and thebest matched synthetic spectrum (dash lines). Lower right panel: The matching of Mg I b linesbetween the degraded FIES spectrum (solid lines) and the best matched synthetic spectrum (dashlines).caused by the uncertainty of the analysis and thus may not be their intrinsic scatters. Since the scatters in thefour GCs are not significant and comparable with the uncertainty of our [Mg/Fe] determination, the derived[Mg/Fe] ratios for member stars of these GCs can testify the application of our method on SDSS data.The member lists of these GCs from SDSS were given by Smolinski et al. (2011), and we select memberstars with high quality spectra of
S/N Mg > . For these stars, micro-turbulences are not available, andwe adopt Equation 1 for log g > g < T eff and ξ t data using UVES giant and sub-giant samples (Lindet al. 2009). The uncertainty of the micro-turbulence is about 0.3 kms − , which only has a negligible effecton the derived [Mg/Fe] ratio as discussed in Sect. 3.2 and also shown in Figure 3. We apply the line-profile-matching method to the member stars of the four GCs, and obtain the mean [Mg/Fe] and its scatter for eachcluster by averaging individual values of member stars. The detailed procedure for the determination of[Mg/Fe] from the SDSS spectra will be described in Sect. 4. ξ t = 1 .
01 + 4 . × − ( T eff − . × − ( T eff − (1) X. Li et al. -0.04-0.02 0 0.02 0.04 0.06 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
Difference of [Mg/Fe] [Mg/Fe] of paper turb-0.3trub+0.3
Fig. 3
The deviations of [Mg/Fe] ratios by varying ξ t with ± − for 47 stars in NS10. Table 2
Stellar parameters and [Mg/Fe] ratios of four globularclusters by HRS analysis
Cluster Num(stars) [Fe/H] σ [ F e/H ] [Mg/Fe] σ [ Mg/F e ] Reference a M92 6 -2.31 0.08 +0.19 0.19 S96M13 25 -1.50 0.05 +0.24 0.15 C05M3 13 -1.39 0.05 +0.40 0.12 C05M71 24 -0.78 0.10 +0.36 0.09 RC02
Notes: a S96:Shetrone 1996; C05:Cohen & Mel´endez 2005; RC02:Ram´ırez & Cohen 2002 ξ t = 8 . − . × − T eff + 1 . × − ( T eff ) (2)Stellar parameters and mean [Mg/Fe] ratios of the four clusters are adopted from the literatures whichare based on high-resolution abundance analysis as listed in Table 2. The comparisons of [Mg/Fe] from ourwork and HRS studies are shown in Figure 4. Although the scatter in [Mg/Fe] of the four GCs is about 0.12 dex , there is a systematic shift to a higher value than previous work by about . dex . We suspect that thedifference between atmospheric parameters derived from SSPP and those by HRS studies could explain theobserved systematic shift. Figure 5 shows the distribution of the T eff versus log g of the GCs’ member starscompared with Dartmouth isochrones with metallicities and age close to those of each GC. It is shown thatfor M13, M3 and M71, the HRS sample nicely follow the isochrones, while our samples show a noticeabledeviation, with + dex in log g and −
250 K in T eff . We thus re-calculate the [Mg/Fe] ratios of samplestars of these GCs by varying log g by + dex and T eff by −
250 K, and find a difference in [Mg/Fe] of0.26 dex and 0.2 dex , respectively, which is able to explain the shift in [Mg/Fe]. We also take this effectinto account in the following selection of strongly Mg-enhanced stars. http://stellar.dartmouth.edu/models/index.html Search for Strongly Mg-enhanced Stars from Sloan Digital Sky Survey 9 -0.4-0.2 0 0.2 0.4 0.6 0.8 1 1.2-2.6 -2.4 -2.2 -2 -1.8 -1.6 -1.4 -1.2 -1 -0.8 -0.6 [Mg/Fe] [Fe/H]
M92M13M3M71my-workprevious-paper
Fig. 4
The comparison of [Mg/Fe] of four globular clusters between our work and HRS analysis.The mean [Mg/Fe] and rms are 0.47 ± ± ± ± In order to make a direct comparison, we apply our method on SDSS spectra for 136 extremely metal-poorstars which has been observed by the High Dispersion Spectrograph of the Subaru Telescope and analysedin details by Aoki et al. (2013) (hereafter Aoki13). Our derived [Mg/Fe] ratios are then compared with thosefrom Aoki13. Among the 136 stars, 45 stars are excluded due to low S/N Mg ( <
30) and another 12 stars areexcluded due to their weak Mg I b feature or bad continuum from the blended C band. A final sample forcomparison includes 80 stars, resulting in an average deviation of 0.21 dex between our work and Aoki13with a scatter of 0.24 dex . We check the distribution of these 80 stars on the diagram of T eff versus log g , and find 19 stars located outside one sigma region of the theoretical isochrone, which are consideredwith unreliable atmosphere parameters and then excluded from the comparison sample. The remaining 61stars with S/N Mg >
30 and reliable atmospheric parameters can be used to make a star-to-star comparisonof [Mg/Fe]. The left panel of Figure 6 shows a good agreement between the two sets of data. There isno obvious deviation from the one-to-one line, and the scatter of 0.18 dex is within the typical error ofour measurements. Note that for the 61 stars, the atmospheric parameters derived form SSPP (which wehave adopted) are not exactly the same as those by Aoki13, but there should not be large differences afterchecking the locations of these objects in the T eff versus log g diagram. But the spectra of the two worksand the procedure of deriving [Mg/Fe] are very different: our values are derived from low-resolution spectravia the line-profile-matching method, while Aoki13 uses high-resolution spectra and measures the [Mg/Fe]ratio from individual magnesium lines. The consistency in the [Mg/Fe] ratios between our work and Aoki13indicates that our method of deriving [Mg/Fe] from low-resolution SDSS spectra is reliable. In particular,stars with high [Mg/Fe]( > . dex ) in Aoki13 are well reproduced by our method.Out of the 136 stars, there are 122 with [ α /Fe] ratios available from the SDSS DR7 database. The rightpanel of Figure 6 compares the [ α /Fe] derived from SSPP DR7 and [Mg/Fe] from Aoki13. Apparently, Fig. 5
The log g versus T eff diagram of member stars of four GCs with the updated Dartmouthisochrones (2012 version) from Dotter et al. (2008) (solid lines). Stars from HRS analysis areshown in rhombuses and SDSS cluster members in pluses.the agreement is not so good and there is no object with [ α /Fe] larger than 0.4 dex if the SSPP [ α /Fe]is adopted. So indeed, described in Lee et al. (2011), the measured [ α /Fe] from averaging four individual α -elements by using weighting factors may not correctly represent the overall content of specific elementlike Mg. Therefore, the [ α /Fe] from SSPP cannot be used to search for strongly Mg-enhanced stars. The uncertainties of the measured [Mg/Fe] come from the spectra, the models and the errors of atmosphericparameters, which can be quantitatively estimated from a Monte Carlo simulation. We choose two stars,SDSS J134922.91+140736.9 and SDSS J130538.1+194305.6, as examples of our sample. In particular,one is on the turnoff stage and the other is on the red giant branch. Moreover, SDSS J134922+140736 isconfirmed as a strongly Mg-enhanced star (Sbordone et al. 2012). Then we perform the Monte Carlo sim-ulation for the two stars based on 500 sets of atmospheric parameters, which are generated with the errorsof atmospheric parameters of SSPP with ± K , ± . dex and ± . dex for T eff , log g and [Fe/H],respectively. Tests are made and proved that 500 sets of atmospheric parameters are enough for this simu-lation while larger sets of data will not result much differently. The mean value and rms of the simulation is + ± dex for SDSS J134922.91+140736.9, and + ± dex for SDSS J130538.1+194305.6,as shown in Figure 7. It is worth noticing that, the small scatter of 0.18 dex on [Mg/Fe] between our result Search for Strongly Mg-enhanced Stars from Sloan Digital Sky Survey 11
Fig. 6
The comparisons of [Mg/Fe] from our work (left panel) and Lee et al. (2011) (right panel)with the HRS from Aoki et al. (2013). Left panel: 12 stars (rhombuses) are excluded due to theirweak Mg I b feature or bad continuum from the blended C band of SDSS spectra, and 19 stars(triangles) are excluded due to their unreliable atmospheric parameters from SSPP.and Aoki et al. (2013) shows that the external error may not be as large as 0.2 dex , and the internal error of0.3 dex estimated by the Monte Carlo simulation should be the upper limit of the uncertainty. The determination of [Mg/Fe] ratios from SDSS spectra is carried out by the following two steps. Firstly, wetransform vacuum-based SDSS spectra into the air-based scale and shift the spectra to the rest frame withthe radial velocity of SSPP. Secondly, we select 12 continuum windows in the wavelength range of 4900 ˚A -5400 ˚A based on a set of high-resolution spectra provided by the ELODIE archive with stellar atmosphericparameters covering the similar range to that of our sample. We obtain the continuum by a polynomial fitto these windows of the whole wavelength range of 4900 ˚A - 5400 ˚A , and normalize the SDSS spectraby dividing the spectra with the continuum. Finally, the synthetic spectra are normalized in the same wayto ensure the observed SDSS spectra and the synthetic spectra can be well matched. Following the methodand fixed parameters described in Sect. 3.2, the [Mg/Fe] ratios are derived for our 14,850 sample stars andtheir distributions of [Mg/Fe] versus [Fe/H] are shown in Figure 8.Among our sample, 174 stars are with [Mg/Fe] greater than 1.0 and located above the dotted line inFigure 8. Individual spectra of the 174 stars are checked by eyes and stars with unclear Mg I b feature ornot-well-defined continuum are excluded. Finally, 84 strongly Mg-enhanced candidates are selected throughthe interactive checking shown with asterisks in Figure 8. In particular, SDSS J134922+140736 ,the stronglyMg-enhanced star discovered by Sbordone et al. (2012) is included in our candidate list; however, SDSSJ084016+540526 discovered by Aoki et al. (2013) is not selected into our sample due to its low S/N Mg ( <
30) in the SDSS spectra. Note that the high [Mg/Fe] of these candidates may be incorrectly estimatedif their atmospheric parameters from SSPP are unreliable. Therefore, in order to check their atmosphericparameters ,we compare the 84 candidates in the T eff versus log g diagram with isochrones from Dotter Fig. 7
The Monte Carlo simulation of the [Mg/Fe] distribution for SDSS J134922+140736(T eff =6342 K, log g =3.89, [Fe/H]= − + ± eff =5230.05 K, log g =2.46, [Fe/H]= − + ± dex ,while our selection criterion of [Mg/Fe] > + ∼ + dex for most EMP stars, we could still expect type B to be qualified candidates of strongly Mg-enhanced stars.It is known that the strength of Mg I b lines are sensitive to the log g and if the log g is underestimated,the Mg abundance will be overestimated. Besides, most of members of the GCs used for calibration insection 3.3 are giants and subgiants, while most of our 33 candidates are turn-off and dwarf stars. Therefore,we have further checked the effect of the assumed underestimation of log g on the derived [Mg/Fe] ratiosof the final 33 candidates. According to Figure 5, it seems SSPP underestimated log g by 0.5 dex which isadded to the log g of the 33 candidates. As shown in Figure 10, the effects of increase of log g on [Mg/Fe]of turn-off and dwarf stars are slightly larger than those of giants and sub-giants in the order of 0.03 dex .For the turn-off and dwarf stars, the average decrease of the [Mg/Fe] to the assumed increase of log g is0.28 dex . If such an decrease in [Mg/Fe] is taken into account, the fractions of [Mg/Fe] greater than 1.0, Search for Strongly Mg-enhanced Stars from Sloan Digital Sky Survey 13
Fig. 8
The [Fe/H] versus [Mg/Fe] diagram of the sample stars. Asterisks are 84 strongly Mg-enhanced candidates after interactive checking of clear Mg I b line profile and well-defined con-tinuum. The red squares are the 33 strongly Mg-enhanced candidates after atmospheric parame-ters check.0.9, 0.8 and 0.7 of our 33 candidates are 34%, 49%, 93% and 100%, respectively. Even under the extremecase of underestimation of log g by 0.5 dex , the [Mg/Fe] of all the 33 candidates are still obviously largerthan those of normal stars and hence our selection is confirmed to be reliable. Based on a line-profile-matching method applied on the Mg I b feature of SDSS spectra for a sample of14,850 F and G stars, 33 strongly Mg-enhanced candidates with [Mg/Fe] > dex , indicating the validationof our method. Moreover, we derive [Mg/Fe] from the Mg I b feature for member stars of four globularcluster observed in SDSS and they are consistent with those from previous HRS investigations. The methodpresented here can be used to the spectral analysis of the LAMOST (The Large Sky Area Multi-ObjectFiber Spectroscopic Telescope) Galactic survey. A larger spectroscopical survey of the Galactic stars (Zhaoet al. 2012) will provide higher probability to find more candidates as such. Meanwhile, future follow-uphigh-resolution spectroscopic observations of the 33 candidates are of great interest to confirm their high Fig. 9
The log g versus T eff diagram for four different metallicities [Fe/H] ( − − . , − . and − . ) with a age coverage of 5 Gyrs:(a) [Fe/H] = − . ∼ − . , t=11 ∼
15 Gyrs; (b) [Fe/H]= − . ∼ − . , t=11 ∼
15 Gyrs; (c) [Fe/H] = − . ∼ − . , t=11 ∼
15 Gyrs; (d) [Fe/H]= − . ∼ − . , t=11 ∼
15 Gyrs. The isochrones corresponding to (a) − (d) are shown in solidlines, and the dash line are the isochrones with a variation in T eff and log g of 250 K and 0.5 dex , respectively. Fig. 10
Variations of [Mg/Fe] stem from 0.5 dex increase on log g , Δ [Mg/Fe]=[Mg/Fe] logg -[Mg/Fe] ( logg +0 . . Search for Strongly Mg-enhanced Stars from Sloan Digital Sky Survey 15
Table 3
Catalog of strongly Mg-enhanced candidates
Star Plate MJD Fiberid T eff ( K ) [Fe/H] log g RV (km/s) S/N [Mg/Fe] TypeSDSS J001820.36-091833.0 652 52138 545 6334.59 -2.61 3.67 -57.71 51.33 1.11 ASDSS J012552.41+382358.4 2063 53359 130 6538.28 -2.34 3.89 -352.44 75.78 1.25 ∗ ASDSS J025432.96+354104.5 2378 53759 408 6400.75 -2.53 3.65 -274.01 56.12 1.18 ∗ ASDSS J084016.16+540526.4 2316 53757 515 6289.86 -2.54 3.69 -9.63 34.44 1.05 ASDSS J085650.28+401730.9 1199 52703 437 6246.56 -2.34 3.77 -28.42 47.48 1.17 ASDSS J094649.03+145432.5 2582 54139 407 6289.79 -2.60 3.91 126.12 41.25 1.15 ASDSS J105018.63+000049.6 2389 54213 556 5233.95 -2.67 2.28 312.59 38.39 1.10 ASDSS J110821.68+174746.5 2491 53855 389 6144.90 -2.91 3.61 -16.65 48.54 1.22 ∗ ASDSS J120231.14+204922.4 2893 54552 340 6388.34 -2.07 3.99 147.55 47.46 1.06 ASDSS J125422.99+202619.5 2899 54568 332 6407.53 -2.70 3.73 311.96 53.37 1.19 ASDSS J125712.60+592129.0 2446 54571 110 6393.55 -2.16 3.91 35.83 52.11 1.15 ASDSS J130047.06+601828.3 2446 54571 626 6586.67 -2.46 3.88 -199.92 53.60 1.11 ASDSS J130538.01+194305.6 3235 54880 194 5230.05 -2.25 2.46 -100.78 54.67 1.00 ASDSS J134922.91+140736.9 1777 53857 479 6342.81 -2.83 3.89 -75.47 39.29 1.12 ASDSS J140038.27+230515.2 2784 54529 464 6431.97 -2.07 3.74 -39.70 50.77 1.12 ASDSS J140501.51+361759.9 2906 54577 307 6505.97 -2.50 3.93 -73.52 54.84 1.22 ∗ ASDSS J154120.53+085602.7 1724 53859 420 6036.23 -2.51 4.20 25.75 40.86 1.02 ASDSS J164023.94+233349.5 1571 53174 617 6296.61 -2.53 3.70 -98.69 43.01 1.24 ∗ ASDSS J172813.66+081011.7 2797 54616 477 6417.26 -2.16 3.83 -187.82 46.85 1.07 ASDSS J233534.77+094331.5 2628 54326 380 6542.45 -2.47 3.77 36.89 59.74 1.26 ∗ ASDSS J014419.25-084818.7 2816 54400 596 6592.59 -2.30 3.55 -106.03 67.45 1.16 BSDSS J015505.42-000421.1 2851 54485 356 5234.59 -2.63 1.61 -176.20 37.35 1.32 ∗ BSDSS J084444.69+063124.0 2317 54152 349 6359.19 -2.57 3.51 224.77 46.30 1.17 ∗ BSDSS J120624.14+184411.2 2893 54552 129 6422.26 -2.91 3.47 -123.45 84.56 1.30 ∗ BSDSS J123850.76+173155.3 2599 54234 554 6466.74 -2.19 3.64 17.74 48.53 1.04 BSDSS J131654.14+391830.4 3240 54883 351 6257.38 -2.07 4.10 62.58 46.86 1.07 BSDSS J135718.30+194052.9 2770 54510 210 6504.55 -2.53 4.09 -77.95 54.28 1.15 BSDSS J161021.87+171130.1 2177 54557 382 6409.45 -2.28 3.59 -76.88 58.79 1.09 BSDSS J172229.03+270858.8 2182 53905 429 6058.25 -2.52 3.98 -79.94 52.19 1.09 BSDSS J172556.84+081101.8 2797 54616 383 5010.62 -2.97 1.54 -355.32 53.48 1.27 ∗ BSDSS J173113.88+334921.2 2253 54551 407 5328.50 -2.85 2.18 -228.16 50.61 1.25 ∗ BSDSS J204224.42-062424.7 1916 53269 243 5905.13 -2.00 3.16 2.08 45.67 1.12 BSDSS J222617.34+010644.9 1144 53238 605 5388.60 -2.43 2.25 -258.87 55.46 1.08 B
Notes: ∗ Even under the extreme case of underestimation of log g by 0.5 dex , the [Mg/Fe] of these candidatesare still larger than 1.0 dex . Mg abundances, and detailed chemical abundances of other elements could help us to understand the originsof these peculiar stars and to further probe the formation and chemical evolution of the Galaxy.
Acknowledgements
XL gives thanks to Lan Zhang, Jing Ren and Wei Wang for their helpful suggestionsand to Andreas Koch for providing us the linearly interpolated program of atmospheric models, especialthanks also go to Poul Erik Nissen for kindly providing of FIES spectra. This work was supported bythe National Natural Science Foundation of China under grant Nos. 11390371, 11233004, 11222326, and11103030.
References
Ahn, C. P., Alexandroff, R., Allende Prieto, C., et al. 2012, ApJS, 203, 21 2, 3Ahn, C. P., Alexandroff, R., Allende Prieto, C., et al. 2014, ApJS, 211, 17 2Allende Prieto, C., Sivarani, T., Beers, T. C., et al. 2008, AJ, 136, 2070 3Aller, L. H., & Greenstein, J. L. 1960, ApJS, 5, 139 1Aoki, W., Ando, H., Honda, S., et al. 2002a, PASJ, 54, 427 2Aoki, W., Beers, T. C., Christlieb, N., et al. 2007a, ApJ, 655, 492 2Aoki, W., Beers, T. C., Lee, Y. S., et al. 2013, AJ, 145, 13 1, 2, 9, 11Aoki, W., Bisterzo, S., Gallino, R., et al. 2006, ApJ, 650, L127 2Aoki, W., Honda, S., Beers, T. C., et al. 2005, ApJ, 632, 611 2Aoki, W., Honda, S., Beers, T. C., et al. 2007b, ApJ, 660, 747 2Aoki, W., Norris, J. E., Ryan, S. G., Beers, T. C., & Ando, H. 2000, ApJ, 536, L97 2Aoki, W., Norris, J. E., Ryan, S. G., Beers, T. C., & Ando, H. 2002b, PASJ, 54, 933 2Bensby, T., Feltzing, S., Lundstr¨om, I., & Ilyin, I. 2005, A&A, 433, 185 2Bruntt, H., Bedding, T. R., Quirion, P.-O., et al. 2010, MNRAS, 405, 1907 7Carretta, E., Bragaglia, A., Gratton, R., & Lucatello, S. 2009, A&A, 505, 139 5Castelli, F., & Kurucz, R. L. 2003, in Modelling of Stellar Atmospheres,
IAU Symposium , vol. 210, editedby N. Piskunov, W. W. Weiss, & D. F. Gray, 20P 4Cohen, J. G., Christlieb, N., McWilliam, A., Shectman, S., & Thompson, I. 2011, in RR Lyrae Stars, Metal-Poor Stars, and the Galaxy, edited by A. McWilliam, 239 2Cohen, J. G., Christlieb, N., Thompson, I., et al. 2013, ApJ, 778, 56 2Cohen, J. G., McWilliam, A., Shectman, S., et al. 2006, AJ, 132, 137 2Cohen, J. G., & Mel´endez, J. 2005, AJ, 129, 303 8Dotter, A., Chaboyer, B., Jevremovi´c, D., et al. 2008, ApJS, 178, 89 10, 11Gehren, T., Liang, Y. C., Shi, J. R., Zhang, H. W., & Zhao, G. 2004, A&A, 413, 1045 2Gehren, T., Shi, J. R., Zhang, H. W., Zhao, G., & Korn, A. J. 2006, A&A, 451, 1065 4
Search for Strongly Mg-enhanced Stars from Sloan Digital Sky Survey 17
Goriely, S., & Siess, L. 2005, in From Lithium to Uranium: Elemental Tracers of Early Cosmic Evolution,
IAU Symposium , vol. 228, edited by V. Hill, P. Francois, & F. Primas, 451–460 2Karakas, A. I., & Lattanzio, J. C. 2003, PASA, 20, 279 2Lambert, D. L., Heath, J. E., Lemke, M., & Drake, J. 1996, ApJS, 103, 183 4Lee, Y. S., Beers, T. C., Allende Prieto, C., et al. 2011, AJ, 141, 90 3, 4, 10, 11Lee, Y. S., Beers, T. C., Sivarani, T., et al. 2008a, AJ, 136, 2022 3Lee, Y. S., Beers, T. C., Sivarani, T., et al. 2008b, AJ, 136, 2050 3Lind, K., Primas, F., Charbonnel, C., Grundahl, F., & Asplund, M. 2009, A&A, 503, 545 7Masseron, T., Johnson, J. A., Plez, B., et al. 2010, A&A, 509, A93 2McWilliam, A. 1997, ARA&A, 35, 503 2Nissen, P. E., & Schuster, W. J. 2010, VizieR Online Data Catalog, 351, 19010 1, 3Ram´ırez, S. V., & Cohen, J. G. 2002, AJ, 123, 3277 8Reddy, B. E., Lambert, D. L., & Allende Prieto, C. 2006, MNRAS, 367, 1329 2Ryan, S. G., Aoki, W., Norris, J. E., & Beers, T. C. 2005, ApJ, 635, 349 2Sbordone, L., Caffau, E., & Bonifacio, P. 2012, in American Institute of Physics Conference Series,