Detailed abundances in a sample of very metal poor stars

Unevolved metal poor stars are the witness of the early evolution of the Galaxy. The determination of their detailed chemical composition is an important tool to understand the chemical history of our Galaxy. The study of their chemical composition can also be used to constrain the nucleosynthesis of the first generation of supernovae that enriched the interstellar medium. The aim is to observe a sample of extremely metal poor stars (EMP stars) candidates selected from SDSS DR12 release and determine their chemical composition. We obtained high resolution spectra of a sample of five stars using HDS on Subaru telescope and used standard 1D models to compute the abundances. The stars we analysed have a metallicity [Fe/H] between -3.50 dex and -4.25 dex . We confirm that the five metal poor candidates selected from low resolution spectra are very metal poor. We present, the discovery of a new ultra metal-poor star (UMP star) with a metallicity of [Fe/H]= -4.25 dex (SDSS~J1050032.34$-$241009.7). We measured in this star an upper limit of lithium ( log(Li/H)<= 2.0. We found that the 4 most metal poor stars of our sample have a lower lithium abundance than the Spite plateau lithium value. We obtain upper limits for carbon in the sample of stars. None of them belong to the high carbon band. We measured abundances of Mg and Ca in most of the stars and found three new alpha-poor stars.


Introduction
Metal poor stars are the witnesses of the early evolution of our Galaxy. They provide important clues to the formation of the first objects in the Universe. The detailed abundance analysis of their atmosphere reveal observational details that can be compared to theoretical models for nucleosynthesis in the first metal poor massive stars exploding as supernovae. The trends of their abundance ratios as a function of the metallicity can be used to trace the chemical history of our Galaxy. A thorough discussion of the main scientific goals of the study of metal poor stars can be found in several review articles (see for example Frebel & Norris 2015).
First misidentified as early type stars (e.g. Adams et al. 1935), the metal poor stars revealed their true nature as old metal-poor stars only through high resolution spectroscopic analysis (Chamberlain & Aller 1951). The catalogue of high velocity stars of Roman(1955) was extensively use as source of low metallicity stars (e.g. Greenstein et al. 1957;Wallerstein & Helfer 1959;Wallerstein 1962;Wallerstein et al. 1963). Although this field of research was largely dominated by the work of astronomers in the United States, a growing inter-⋆ Based on data collected at Subaru Telescope, which is operated by the National Astronomical Observatory of Japan. est began also in Europe (e.g Baschek 1963;Cayrel & Fringant 1964).
The search for the most metal poor stars has made great progresses with the survey work of Beers et al. (1985) and the discovery of several very metal poor stars with [Fe/H] ≤ -3.00 dex (Norris et al. 1993). Their survey was making use of a temperature index based on the H β hydrogen absorption line and a metallicity index from the strong Ca H&K lines which can be detected and measured on low resolution spectra even at low metallicity. The high resolution follow-up observations of metal poor candidates led to a large number of publications among them the series of articles from the First Stars ESO Large programme  published by the Cayrel's group (Bonifacio et al. 2009;Cayrel et al. 2004, and references therein). They also showed how rare are the very metal poor stars as revealed from the metallicity distribution function, implying larger and and deeper survey to find new very metal poor candidates. The Hamburg ESO survey whose first aim was the search for new quasars has also been very successful in the search and the discovery of several stars with [Fe/H] with metallicity lower than -4.00 dex , and references therein). More recently, the Sloan Digital Sky Survey (SDSS) spectroscopic survey has been used to identify new metal poor candidates (Helmi et al. 2003). Ludwig et al. A&A proofs: manuscript no. EMP_Subaru_V5 (2008) have developed an analysis tool that allows us to estimate the metallicity of Turn Off (TO) stars from the low resolution SDSS spectra. Follow up observations on large telescope has been used to confirm their metallicity and measure abundance ratios. This detection method has been quite successful and led to the discovery of several interesting very metal poor stars (Caffau et al. 2011a(Caffau et al. ,b, 2012(Caffau et al. , 2014Bonifacio et al. 2015;Caffau et al. 2016;Bonifacio et al. 2018;François et al. 2018). Other surveys have also been used to detect metal poor stars, as the Apache Point Observatory Galactic Evolution Experiment (APOGEE, Majewski et al. (2016)), a survey toward the galactic centre, where the stellar density allows large multiplex spectroscopic observations. We could also mention the Radial Velocity Experiment (RAVE) survey which first aim is the study of the galactic dynamics from a radial velocity census of stars (Steinmetz et al. 2006), the Skymapper Sky Survey (Keller et al. 2007;Wolf et al. 2018) performing wide field imaging in five wide bands and a narrow band centred on Ca H&K absorption lines using the Skymapper telescope. The photometric SkyMapper Southern Sky Survey (Keller et al. 2007) discovered two of the most extremely ironpoor stars: SMSS J031300.36-670839.3 (Keller et al. 2014) and SMSS J160540.18-144323.1 (Nordlander et al. 2019). More recently, the PRISTINE survey (Starkenburg et al. 2017) based on Canada France Hawaii telescope (CFHT) large field imaging uses a dedicated narrow band filter centred on Ca H&K absorption lines, combined with SDSS broad-band g and i photometry.
The common point between these different sources of metal poor stars candidates, is that they require high resolution spectroscopic follow-up observations to confirm their low metallicity and determine their detailed chemical composition.
From the recent analysis of SDSS DR12 data, we have detected new extremely metal poor candidates that have never been observed at high resolution. In this article, we report the detailed analysis of five new extremely metal poor candidates observed with the High Dispersion Spectrograph (HDS) installed on the Subaru telescope atop Mauna Kea volcano in Hawaii. Similar observations have been conducted for a second set of stars visible from the southern hemisphere using the X-SHOOTER spectrograph installed on the UT2 (KUEYEN) at the ESO very large telescope (VLT) on Cerro Paranal in Chile in the framework of a french-japanese collaboration. The results have been published in François et al. (2018).

Observation
The observations have been carried out with HDS installed on the Subaru telescope (Noguchi et al. 2002). The wavelength coverage goes from 4084 Å to 6892 Å. A binning 2x2 has been adopted leading to a resolution of about 40000. The logbook of the observations is given in table 1 . Standard data reduction procedures were carried out with the IRAF Echelle package 1 . Care has been taken to remove the sky background as most of the exposures were affected by the moon illumination. Fig. 1 shows the spectra of the stars of our sample centred on the magnesium triplet. The continuum level of the four spectra located in the lower part of the plot has been shifted for clarity. The spectra are presented with decreasing metallicity from the top to the bottom of the figure.

Stellar parameters
The stellar parameters have been derived taking into account the SDSS photometry.
The effective temperatures in Table 2 have been computed by Caffau et al. (2013) . The effective temperature has been derived from the photometry, using the (g − z) 0 colour and the calibration described in Ludwig et al. (2008) taking into account the reddening according to the Schlegel et al. (1998) extinction maps and corrected as in Bonifacio et al. (2000). The Gaia parallaxes (Arenou et al. 2018, Gaia Collaboration et al. 2018) for our stars are imprecise. SDSSJ081554.26+472947.5 and SDSSJ091753.19+523004.9 parallaxes have a relative error smaller than ∼ 20%. SDSSJ124304.19−081230.6 and SDSSJ153346.28+155701.8 parallaxes have a relative error of the order of ∼ 50%. One star (SDSSJ105002.34+242109.7 ) has a negative parallax. In order to get some insight in the luminosity of this star, we used the distance estimates of Bailer-Jones et al.  (Bressan et al. 2012) isochrone of metallicity -2.5 dex and an age of 12 Gyrs. The location of the stars clearly reveals that the stars are not giant stars. Indeed, the G 0 for our stars are between 3.0 and 6.5 which correspond to log g ranging from ≃ 3.5 to 4.7 based on the isochrone. It confirms that four of our stars are dwarf stars and one seems slightly evolved. Adopting log g = 4.00 for the gravity of our stars is suited for our analysis. We remind that the selection of the stars is based on the dereddened (g − z) and (u − g) colours: 0.18 ≤ (g − z) 0 ≤ 0.70 and (u − g) 0 ≥ 0.70. As discussed in Bonifacio et al. (2012), this selects the stars of the halo turn-off and excludes the majority of the white dwarf stars. A microturbulent velocity fo 1.5 km/s suitable for stars with log g =4 dex has been adopted following the results of Barklem et al. (2005). The metallicities shown in Table 2 have been computed by Caffau et al. (2013) using the code MyGisFoS (Sbordone et al. 2010) and the SDSS spectra of the stars.
From the radial velocities given in Table 2 and using the publicly licensed code GalPot 2 , which is described by Dehnen & Binney (1998), we computed the kinematic properties of our sample of stars we have analysed and some of the quantities derived from their Galactic orbits. The results are shown in Table 3. In this table, L z is the angular momentum, R is the galactocentric radius (cylindrical), R min and R max are respectively the minimum and the maximum values of the galactocentric radius (cylindrical) of the orbit, E is the Energy of the orbit. Z max is the maximum galactocentric height of the orbit. The space velocities (U, V, W) are with respect to the Local Standard of Rest, U is positive towards the Galactic anti-centre,V in the direction of the Galactic rotation and W is perpendicular to the Galactic plane, positive in the northern Galactic hemisphere.  1. Parts of the HDS Subaru spectra centred on the magnesium triplet. The continuum of the four spectra located in the lower part of the plot has been shifted downwards for clarity. The two lower spectra reveal the presence of residuals of the sky spectrum in the region 5168-5170 Å and 5185-5186 Å We list also the mean specific angular momentum (angular momentum per unit mass) for the stars along their orbits, in units of kpc × km s −1 . In Fig. 3 the Toomree diagram and the orbital characteristics of our sample are presented. 4 of the stars have large eccentricities and high values of Z max indicating that they are halo stars. The remaining star with the low Z max =2.81 Kpc has an eccentricity of 0.82 and also likely a halo star.

Analysis
We carried out a classical 1D LTE analysis using OS-MARCS model atmospheres (Gustafsson et al. 1975(Gustafsson et al. , 2003(Gustafsson et al. , 2008Plez et al. 1992;Edvardsson et al. 1993). The abundances used in the model atmospheres were solar-scaled with respect to the Grevesse & Sauval (2000) solar abundances, except for the α-elements that are enhanced by 0.4 A&A proofs: manuscript no. EMP_Subaru_V5  (Bressan et al. 2012) isochrone of metallicity -2.5 and age 12 Gyr on a G 0 magnitude versus (BP − RP) 0 magnitude diagram. The isochrone on the plot has been computed for an age of 12 Gyr and a metallicity of -2.5 dex. BP-RP (BP and RP are the magnitudes measured respectively by the two low resolution spectrographs, the Blue Photometer (BP) and the Red Photometer (RP) onboard the Gaia satellite) versus the absolute G magnitude. The red symbols represent our stars. Table 2. Adopted stellar parameters for the list of targets. The last column gives the measured radial velocity of the stars after correction from the barycentric velocity. Object T    The abundance analysis was performed using the LTE spectral line analysis code turbospectrum (Alvarez & Plez 1998;Plez 2012), which treats scattering in detail. The carbon abundance was determined by fitting the CH band near to 430 nm (G band). The molecular data that correspond to the CH band are described in Hill et al. (2002). The abundances have been determined by matching a synthetic spectrum centred on each line of interest to the observed spectrum. Table 4 gathers the list of lines which have been used to measure the abundances or evaluate upper limits in our sample of stars. Table 5 lists the computed errors in the elemental abundances ratios due to typical uncertainties in the stellar parameters. The errors were estimated varying T e f f by ± 100 K, log g by ± 0.5 dex and v t by ± 0.5 dex in the model atmosphere of SDSS J091934.08+524014.0, other stars give similar results. In this star, we could measure the Mg, Ca and set limits for Li, Sr, and Ba abundances. The main uncertainty comes from the error in the placement of the continuum when the synthetic line profiles are matched to the observed spectra. In particular, residuals from the sky subtraction may lead to a decrease of the S/N ratio. As the final spectra are build from the the combination of several exposure taken at different epochs hence different barycentric velocities, the features sky residuals are smoothed and degrade the S/N of the spectra. This error is of the order of 0.1 to 0.2 depending on the S/N ratio of the spectrum and the species under consideration, the largest value being for the neutron capture elements. When several lines are available, the typical line to line scatter for a given elements is 0.1 to 0.2 dex.

Results and discussion
The abundance and upper limit results for the sample of stars of this programme have been gathered in Tables 6. For lithium and carbon, the log(X/H) is given whereas [X/Fe] results are presented for magnesium, calcium, strontium and barium. For the star SDSS J081554.26+472947.5, the abundance ratios are given assuming [Fe/H] < -4.10 dex. This upper limit of the [Fe/H] has been derived using the strongest FeI line available in our spectrum.
Normal carbon stars are represented as black circles. Low and high carbon bands stars are shown as grey circles. The classification of CEMP stars follows the scheme from Bonifacio et al. Different symbols have been chosen for CEMP and non-CEMP to show that the high-carbon band CEMP stars are preferentially Li-depleted whereas the lithium abundances in the lowcarbon band are indistinguishable from that of the carbon normal stars. These measurements are consistent with the hypothesis suggested by Bonifacio et al. (2018) that the high carbon CEMP stars are the result of mass transfer from an AGB companion, also suggested by the works of Koch et al. (2011) and Monaco et al. (2012).
We evaluated the upper limit of the lithium abundance in all the stars of our sample. 4 of them reveal a lithium abundance lower than the Spite plateau value (Spite & Spite 1982;Bonifacio et al. 2007;Sbordone et al. 2010). These stars have metallicities below [Fe/H] =-3.7 dex that place them in the region where the "meltdown" of the lithium plateau appears (Sbordone et al. 2010;Aoki et al. 2009;Bonifacio et al. 2007). The star SDSS J081554.26+472947.5, identified as an EMP star by (Carbon et al. 2017;Aguado et al. 2018), has been studied in detail by González Hernández et al. (2020) who found a metallicity of [Fe/H] =-5.49 dex while we found an upper limit of -4.10 dex . It is interesting to note that, by adding more data in this metallicity range, the decrease of the lithium abundance is not constant at a given metallicity. In particular, two stars analysed  Fig. 4, it seems that the Spite plateau appears as an upper limit for the lithium abundance in metal poor unevolved stars. As the metallicity decreases below [Fe/H] =-3.0 dex, the dispersion of the lithium abundance at a given metallicity seems to increase with decreasing metallicity.

Carbon
In Fig. 5, we plotted the abundance of carbon as a function of [Fe/H]. The results are represented as red squares with down arrows that indicate that the abundances are upper limits. Open red symbols represent the stars for which we found with low [ncapture/Fe] upper limit abundances. We have also added literature results Behara et al. 2010;Cohen et al. 2013;Frebel et al. 2005Frebel et al. , 2006Li et al. 2015;Masseron et al. 2010;Sivarani et al. 2006;Thompson et al. 2008;Yong et al. 2013) . The upper limits of the carbon abundance we found in our five stars are com-patible with them being moderately enhanced in C or C normal. As our measures are upper limits, the stars could be CEMP stars belonging tot the low C band or just C-normal stars .

α and neutron-capture elements
In Fig. 6   Open red symbols represent the stars for which we found with low [n-capture/Fe] upper limit abundances. Open and filled red circles are data published by our group. Other stars from the literature (Sivarani et al. 2006;Frebel et al. 2005Frebel et al. , 2006Thompson et al. 2008;Aoki et al. 2008;Behara et al. 2010;Masseron et al. 2010;Yong et al. 2013;Cohen et al. 2013;Li et al. 2015) are represented as blue squares. Open blue squares are carbon abundances taking with 3D corrections from Gallagher et al. (2016). Black circles filled in red are stars from Bonifacio et al. (2018). The pink symbol represents SDSS J102915+17292, the normal carbon ultra metal poor star discovered by Caffau et al. (2011a). The other symbols are literature data. The black and yellow dashed lines delimit the low-carbon band. Details can be found in Bonifacio et al. (2018).
A&A proofs: manuscript no. EMP_Subaru_V5 the one used by González Hernández et al. (2020) with a difference of 15K on the temperature, a difference of 0.6 dex in log g that has not a strong effect (typically 0.1 to 0.15 dex on the neutral Ca and Mg) on the determination of the abundance of neutral species and the same micro-turbulent velocity. Concerning calcium, we found three stars with sub-solar values, confirming the existence of stars with low [α/Fe] ratios as already suggested by Bonifacio et al. (2018). However, we remind that in very metal-poor stars, under the LTE hypothesis, the resonance line that we used to determine the abundance of calcium leads to an underestimation of the calcium abundance (Spite et al. 2012, and references therein). For turnoff stars, the amplitude of the effect is rather small. Spite et al. (2012) have computed a correction of the order of +0.1 dex in a turnoff star with [Fe/H] = -3.2 dex. In Fig. 7 Aoki et al. 2006). New observations with better S/N ratios would be very interesting to firmly determine the abundance ratios found in these stars.

Conclusions
In this article, we reported the chemical analysis of five extremely metal poor candidates observed with the high dispersion spectrograph (HDS) at the SUBARU telescope. We discover a new UMP stars, SDSS J105002.34+242109.7 with [Fe/H] =-4.25 dex. We could determine the abundances of some elements (C, Mg, Ca, Sr and Ba) in the majority of these stars. The five stars of the sample show abundance ratios which are typical of metal-poor stars in the metallicity range -4.25 dex ≤ [Fe/H] ≤ -3.5 dex. These results show that the method developed by Ludwig et al. (2008) to estimate the metallicity of unevolved stars from low resolution spectra is very efficient. We could measure an upper limit of the lithium abundance in the five stars of our sample. The four most metal poor stars, with a metallicity ranging from -3. to -4.25 dex show lithium abundances below the Spite plateau. Some stars of our sample show a low [α/Fe] content, a characteristic already found in previous studies (Bonifacio et al. 2018, an reference therein).     Roederer et al. (2014). Blue circles : main sequence stars from Roederer et al. (2014). The blue rectangle represents HE1327-2326 a star with an exceptional high [Sr/Fe] ratio Aoki et al. 2006).