Chemical Compositions of Four Metal‐poor Giant Stars

Chemical compositions of metal‐poor stars have long been used to probe the history of the early Galaxy. As the number of very metal‐poor stars having a well‐determined chemical composition has increased, it has become apparent that the metallicity, usually represented by the iron abundance [Fe/H], is not sufficient to predict the abundances of other elements; a real star‐to‐star scatter in abundance ratios appears for many elements among stars more metal‐poor than [Fe/H] of about −2. For stars in the [Fe/H] range −2 to +0.3, the various ratios show almost no "cosmic" scatter. In the light of the cosmic scatter shown by very metal‐poor stars, it is important to analyze as large a sample of such stars as possible, in order to characterize fully the extent of the scatter. This paper represents a modest contribution to that end by presenting abundance analyses of four K giants with [Fe/H ] ∼ -2.4 to −3.1 and by describing an as yet unsuccessful search for late‐type metal‐poor stars in a list of stars with high tangential velocity.

A program of medium‐resolution spectroscopy was undertaken to identify metal‐poor field stars from Lee's (1984) list of stars with tangential velocities estimated to exceed 100 km s⁻¹. Lee compiled his lists from Lowell proper motions (Giclas, Burnham, & Thomas 1971) and an estimate of the parallax (trigonometric or spectroscopic). We elected to concentrate on stars of spectral type K because major surveys of metal‐poor stars have largely been restricted to earlier spectral types (cf. Carney et al. 1996). Stars of spectral type K provide opportunities to measure aspects of a star's chemical composition not readily determinable from warmer stars. The stars observed with medium resolution are presented in Table 1. Estimates of the B or V magnitude and spectral type are taken from SIMBAD. The radial velocity is measured from our spectra (see below). In addition to stars from Lee's sample, we observed three stars from the HK Proper Motion Survey undertaken by Beers, Preston, & Schectman (1985, 1992).

The Cassegrain spectrograph at the 2.3 m Vainu Bappu Telescope at Kavalur (Prabhu, Anupama, & Surendiranath 1998) was used for the spectroscopic survey. With the chosen grating of 1200 grooves mm⁻¹ and a Tektronix 1024 × 1024 24 μm pixel CCD, the intermediate dispersion was 1.3 Å pixel⁻¹. The bright K subgiant (spectral type K1 IV) star HR 4182 was observed as radial velocity standard and also as a near–solar‐metallicity representative.

These spectra enabled us to measure radial velocities (Table 1) to an accuracy of about ±18 km s⁻¹ using the cross‐correlation task FXCORR contained in IRAF software. Our velocities are in fair agreement with the few previously published values located through SIMBAD (see footnotes to Table 1). From inspection of the spectra, it was obvious that five stars from Table 1 are extremely metal‐poor: CS 22166‐16, 22169‐35, 22877‐1, 22877‐11, and 22877‐51. A metal deficiency for these stars was expected from the preliminary estimates of metallicity assigned by Beers et al. (1992). CS 22877‐11 was the subject of a detailed abundance analysis reported by McWilliam et al. (1995) who gave the iron deficiency as [Fe/H ] = -2.9. Strömgren photometry (Schuster et al. 1996) showed CS 22877‐51 to be metal‐poor ([Fe/H ] = -2.45). Lee's star G53‐24 is given [Fe/H ] = 0.18 by Ryan & Norris (1991).

The three very metal‐poor stars from the Kavalur survey with BS 16085−0050 from Anthony‐Twarog et al. (2000) were selected for an abundance analysis. They were observed with the Apache Point Observatory's 3.5 m telescope and the vacuum‐sealed echelle spectrograph. It uses a 31.6 line mm⁻¹ echelle grating with a prism cross‐disperser. The 2048 × 2048 SITe CCD has 24 μm pixels, resulting in a 2 pixel resolving power near 38,000. Owing to the close spacing of the orders on the CCD, it is necessary to employ nonstandard reduction methods. The most significant difference is the use of a hot star instead of a lamp as a flat. As a measure of the quality of the spectra, we note that the signal‐to‐noise ratio (per pixel) in the continuum near 6700 Å was about 70 (CS 22166‐16), 60 (CS 22169‐35), 75 (CS 22877‐1), and 90 (BS 16085−0050).

Inspection of the high‐resolution spectra confirmed that the observed stars are very metal‐poor. Hydrogen lines appeared normal, with no indication of emission. Our abundance analysis was an entirely standard procedure, as described for example in our papers on the RV Tauri variables (cf. Giridhar, Lambert, & Gonzalez 2000). A 1997 version of the spectrum synthesis code MOOG (Sneden 1973) was used with model atmospheres drawn from the grid computed by Kurucz (1993). Lines of Fe i and Fe ii were used to derive the atmospheric parameters: effective temperature T_eff, surface gravity g, and the microturbulent velocity ξ_t. The requirement that the Fe abundance derived from Fe i lines be independent of excitation potential and equivalent width were used to derive T_eff and ξ_t, respectively. Then, the requirement that Fe i and Fe ii lines return the same abundance was used to derive log g. This use of an ionization equilibrium was verified using the Ti i and Ti ii lines. Given the number of lines and the accuracy of the equivalent widths (5%–10%), we consider the uncertainties for these quite similar stars to be about ±150 K in T_eff, ±0.25 in log g, and ±0.2 km s⁻¹ in ξ_t. The iron abundance is determined to about ±0.2 dex. Final results for the atmospheric parameters are given in Table 2. Inspection of the abundances derived for the individual stars (Table 3) show that the ionization equilibrium of Ti and Fe is satisfied with very similar parameters; the abundance differences from Ti i and Ti ii lines are equal to those from the Fe i and Fe ii lines to within 0.15 dex for all stars.

Lines employed in this analysis are presented in theTable 4. The log gf values are taken from compilations provided by R. E. Luck (1992, private communication) and McWilliam et al. (1995). We derived carbon abundances for CS 22877‐1, CS 22166‐16, and CS 22169‐35 using the CH band at 4310–4330 Å. No CH line could be detected for BS 16085−0050 in spite of its low temperature. We used hyperfine structure parameters given in McWilliam et al. for Mn i lines at 4030, 4033, and 4034 Å to synthesize this region for BS 16085−0050 to derive the Mn abundance. The same could not be done for CS 22877‐1, CS 22166‐16, and CS 22169‐35 as Mn i lines were contaminated by CH lines. Elemental abundances are given in Table 3 as [X/H] with the standard error derived from the line‐to‐line scatter, the number of adopted lines, and [X/Fe]. Uncertainties arising from likely errors in the atmospheric parameters can be assessed from Ryan, Norris, & Beers (1996; Table 3 and entries for HD 122563). In almost all cases, our tabulated standard error is the dominant contributor to the total error; the standard error of the mean is formally smaller than that quoted by the square root of the number of contributing lines.

In the light of published results on the composition of extremely metal‐poor stars, we commence our discussion by comparing and contrasting the four stars with previously analyzed stars. This will be done using the ratio [X/Fe]. Our primary reference works are the surveys by McWilliam et al. (1995) and Ryan et al. (1996) and reviews by McWilliam (1997) and Norris (1999). Variation of [X/Fe] with [Fe/H] is well determined from [Fe/H ] = 0.0 to [Fe/H ] ≃ -2 with remarkably little true (cosmic) scatter as long as normal stars are considered (Lambert 1989; Wheeler, Sneden, & Truran 1989). For [Fe/H ]≤ -2, many relations change shape and slope and may develop a significant cosmic scatter. Our stars will be judged as normal if they fall within the range of [X/Fe] given in the reference works. It should be noted that the analytical tools (models, lines, etc.) used here are very similar to those employed by McWilliam et al. and Ryan et al., and both samples included not only dwarfs but giants similar to our stars. Therefore, systematic errors affecting our [X/Fe] should be similar to those of the reference works. Since our principal goal is to relate our stars to the previously analyzed stars, we do not here concern ourselves with the systematic errors arising from defects in the analytical tools, e.g., the use of local thermodynamic equilibrium (LTE) when non‐LTE effects may be significant.

4.1. CS 22877‐1

This C‐rich star at [Fe/H ] = -2.8 has [X/Fe] firmly within the expected range except for Na (possibly underabundant), Al, and Ti (possibly slightly underabundant). The high C abundance is evident from the great strength of the CH bands: our estimate of [C/Fe ] = +1.8 comes from a spectrum synthesis (Fig. 1). Although this star is C‐rich relative to most extremely metal‐poor stars, its [C/Fe] is matched by other stars (Norris, Ryan, & Beers 1997). The Na abundance, [Na/Fe ] = -0.5, is a little outside the range of −0.3 to +0.4 reported by McWilliam et al., also using the Na D lines. Figure 2 shows the Al i line at 3961 Å. The Al abundance, [Al/Fe ] = +0.2, is within the large range of −1.0 to +0.5 found by McWilliam et al. from the 3961 Å Al i resonance line; the other resonance line at 3944 Å was shown by Arpigny & Magain (1983) to be blended with several CH lines, even in stars where CH was not the prominent feature of the spectrum that it is for CS 22877‐1. Ryan et al.'s analyses of the resonance line, however, gave a rather well‐defined "plateau" at [Al/Fe ] ≃ -0.8; relative to this value (and to points in their [Al/Mg] vs. [Fe/H] plot), CS 22877‐1 is Al‐enriched, but analysis of the 3961 Å line is sensitive to the adopted microturbulence. The origin of the difference between the Al abundances of McWilliam et al. and Ryan et al. is unclear, according to Ryan et al. Clearly, there is a need to probe this discrepancy more deeply because classification of Al abundances for the four stars as "normal" or "abnormal" depends on whether the appropriate set of reference abundances is that offered by McWilliam et al. or that by Ryan et al. We do not attempt to resolve the discrepancy. In Figure 3, we show the star's location in [X/Fe] versus [Fe/H] plots for X=Al, α, where α denotes Mg and Ca. Abundances from McWilliam et al. and Ryan et al. are given as a reference. We chose not to include [Si/Fe] as not many good Si i lines were accessible. The abundances of the three measured elements from Sr to Ba are each within the ranges defined by McWilliam et al. and Ryan et al.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

4.2. CS 22166‐16

4.3. CS 22169‐35

4.4. BS 16085−0050

A somewhat surprising outcome of our initial survey of Lee's high‐velocity stars is the absence of metal‐poor stars. Most stars appear to have an abundance near solar. Given that the selection criterion was a tangential velocity in excess of 100 km s⁻¹, we expected to find a rather metal‐poor sample. This apparent puzzle will be considered when our survey is more complete.

Our modest addition to the number of very metal‐poor stars subjected to an abundance analysis reveals several interesting facets about these stars. Two of the stars are evidently very rich in carbon. This is not an uncommon feature of very metal‐poor stars (Rossi, Beers, & Sneden 1999). At solar metallicities, carbon enrichment of a stellar atmosphere is widely associated with enrichment of s‐process elements and attributed to contamination of the star by mass transfer from an asymptotic giant branch star. This attribution is far less appropriate for very metal‐poor stars. Several C‐enriched stars have been shown not to be binaries (Norris et al. 1997), and in addition the C enrichment is not always coupled with s‐process enrichment. In short, mass transfer is a plausible explanation in only some cases, and in others the C enrichment was likely present in the star's natal cloud. In the case of CS 22877‐1 and probably also CS 22877‐16, the carbon enrichment is not related to the s‐process, but heavy elements point to an r‐process connection. Whether these stars are binaries or not is presently unknown.

Very metal‐poor stars are apparently not a monolithic block with respect to the ratio of α‐elements to iron. BS 16085−0050 is α‐enriched at a level rarely seen previously, but this enrichment is not unprecedented: CS 22949‐37 is similar (McWilliam et al. 1995). At the other extreme, CS 22169‐35 joins a small group of α‐poor very metal‐poor stars. Star‐to‐star spread in elemental abundance ratios with respect to iron are seen in some other elements, particularly the heavy elements, where a considerable dispersion has been noted previously.

The large spread in abundances particularly for α‐elements found for stars with [Fe/H ]< -2.5 indicates that before this metallicity level was attained, the early Galaxy experienced very unusual chemical enrichment. Refinement of supernovae models are perhaps required to explain the uneven yield of α‐elements. Galactic chemical enrichment models including the effect of incomplete mixing might help in explaining the observed abundance peculiarities. Perhaps the principal lesson to be drawn from our small sample is that not everything has yet been gleaned from abundance analyses of very metal‐poor stars. The mine is not yet exhausted of valuable ores.