Abundances in the Metal‐poor Common Proper Motion Star BD +00°2058A

Abundances were derived from measured equivalent widths using an updated version of the LTE analysis package MOOG (Sneden 1973), and the l/H_p = 1.25 model atmospheres of R. L. Kurucz (1992, private communication). While a differential line‐by‐line analysis with respect to the Sun is desirable for accurate abundances, most moderate strength lines in BD +00°2058A are quite strong in the Sun, and this might raise concerns about the treatment of damping in performing a consistent solar normalization of the stellar abundances. For cases where this was a concern (Fe i, Cr i, Mg i), normalization of the mean stellar absolute abundance was carried out using the mean solar abundance derived from different (generally weaker) solar features havinggf‐values consistently determined in the same fashion as for the stellar features. For the remaining species (Fe ii, Ni i, Ti ii, Ca i, Si i, O i, S i, Na i, C i, Ba ii, Y ii, Zr ii), solar normalization was carried out on a line‐by‐line basis using consistently determined solar abundances.

The species and wavelengths of the analyzed features are given in columns (1) and (2) of Table 1. Columns (3) and (4) contain the lower excitation potential and oscillator strength. The gf‐values for Fe i are the reasonably accurate (typically ≲15% or so) measures from O'Brian et al. (1991); for three features (λλ4978, 6336, 6400), gf‐values were taken from Bard & Kock (1994) or Bard, Kock, & Kock (1991, hereafter BKK). Thegf‐values for Fe ii are from Biemont et al. (1991). The remaining oscillator strengths are as follows: mean values of Blackwell, Menon, & Petford (1984) and Tozzi, Brunner, & Huber (1985) for Cr i; both Edvardsson et al. (1993, hereafter E93) and Thévenin (1990) for both Ni i and Si i; mean values from Chang (1990) and Chang & Tang (1990) for Mg i; Wiese, Smith, & Glennon (1966) for O i; the Kurucz (1995) CD‐ROM linelist value for S i; Table 2 of Holweger (1971) for Na i; Table 1 of Grevesse et al. (1991) for C i; the sources utilized and described by King et al. (1998, hereafter K98) for Ti ii and Ca i; and the sources utilized and described by King, Deliyannis, & Boesgaard (1996) for Ba ii, Y ii, and Zr ii. Owing to its significant strength in the Sun, hyperfine structure components of the λ4554 Ba ii were taken into account using the data from McWilliam (1998) and assuming the solar isotopic proportions from Anders & Grevesse (1989).

The stellar equivalent widths and the derived absolute stellar abundances are listed in columns (5) and (6) of Table 1. The solar line strengths measured from a daytime sky spectrum obtained with the same instrumental configuration used to acquire our stellar spectra, and the absolute abundances derived therefrom, are given in columns (7) and (8) of Table 1. When no solar data are listed, this indicates either that solar data for the same features were not used to conduct a line‐by‐line normalization or that we used the solar abundances for the same features from Table 3 of K98, which were derived in a wholly self‐consistent fashion (same atmospheres, damping assumptions, consistent atomic data, etc.) as the present abundances. If the former, then the mean solar abundances of a particular species from K98 were used to normalize the mean stellar abundance derived here. The K98 solar data are from a high‐resolution, high S/N lunar spectrum obtained with the Keck HIRES spectrograph. We searched for any systematic instrumental differences by measuring 16 Fe i lines having a range in strength from a daytime sky spectrum obtained with our McDonald 2.7 m configuration. The comparison is shown in Figure 1 and indicates outstanding agreement. The mean equivalent width difference (Keck minus McDonald) is 0.1 ± 0.5 (s.d. of 1.8) mÅ; in relative terms the differences is 0.6% ± 1.0% (s.d. of 3.9).

Zoom In Zoom Out Reset image size

Peripheral notes concerning the analysis and data are as follows: A typical modest enhancement (2.2 times) in the van der Waals broadening was employed for all species except Ba ii. Following several studies (e.g., the seminal solar analysis of Holweger & Mueller 1974), a slightly larger van der Waals enhancement factor of 3.5 was adopted for the Ba calculations. The K98 solar Mg and Ca data was supplemented with the λ5711 Mg i and λ6455 Ca i features measured in our McDonald sky spectrum and are listed in Table 1; the atomic data for this Ca feature are taken from Tomkin et al. (1997). The stellar measurement of the weak S i line is more uncertain than for the other lines.

The only two Na i lines available to us are not entirely satisfactory. The profile of the λ6160 feature is not clean in our stellar spectrum; the measured line strength is thus uncertain. The solar λ8183 feature is strong, and its abundance is sensitive to damping and the accuracy of the measured line strength. The normalized [Na/H] values are in outstanding agreement, however. For the λ9061 C i feature, we adopted a gf‐value which gave an absolute solar abundance equal to the mean from the other two C i lines. This feature may be affected by modest blending. The equivalent width measured from the Kurucz et al. (1984) atlas is 134 mÅ, 14 mÅ less than the measure from our McDonald sky spectrum. We have retained the latter, however, to mitigate the effects of blending which would likely similarly affect the stellar line too. Accordingly, ourgf‐value is 0.1–0.2 dex larger than the values from Grevesse et al. and the Kurucz (1995) CD‐ROM. Happily, the three lines yield normalized C abundances in fine agreement. Given the significant strength of the λ4554 Ba ii feature, its strength was measured from the high‐resolution, high S/N solar flux atlas of Kurucz et al. (1984).

Initial parameters of T_eff = 6065 K, log g = 4.25, ξ = 1.5 km s⁻¹, and [m/H] = -1.5 were adopted to begin our analysis. The remainder of the analysis, while not described as such, was conducted iteratively such that successive results were used to adjust the initial parameters until self‐consistency was achieved. The weaker of those Fe i lines listed in Table 1 were considered first. Minimizing the trend of derived absolute abundances with excitation potential yielded T_eff = 6130 K. Repeating the analysis with the stronger lines and minimizing the trend of abundances with line strength resulted in an estimate of ξ = 1.4 km s⁻¹. The final satisfactory state of affairs between abundance and excitation potential/line strength is shown in Figure 2, where the Fe i lines have been supplemented with the Fe ii, Ti ii, Cr i, and Mg i results in the abundance versus line strength plane.

Zoom In Zoom Out Reset image size

We find [Fe/H] = -1.16 ± 0.03 (internal mean error) with a 0.10 dex scatter in the absolute Fe i abundances. Accordingly, the final parameters and results were derived with a refined model atmosphere metallicity of [m/H] = -1.0. Ionization balance was used to estimate the final log g value. The assumed model atmosphere gravity was altered until we achieved a mean result of [Fe ii/H] = -1.16, the result from our Fe i lines. This procedure yields log g = 4.19. The per‐line Fe ii standard deviation is a satisfactory 0.085 dex (σ = 0.03 dex), similar to the Fe i value.

The T_eff uncertainty was determined by reanalyzing the Fe i lines with model atmospheres of different T_eff and examining the resulting slope in the abundance‐excitation plane. We find that the 1 σ level changes occur for 65 K deviations and take this as our T_eff uncertainty. Similar examination of slope changes in the abundance–line strength plane after varying the assumed microturbulence suggests a 1 σ level uncertainty of ±0.15 km s⁻¹. The uncertainty in log g from the internal mean abundance errors in the Fe i and Fe ii results is ±0.11 dex. However, the derived gravity is also sensitive to the assumed temperature since this affects the Fe i abundance. Including the above T_eff uncertainty, the final total estimate is σ(log g) = ±0.19 dex.

The final normalized abundance ratios (i.e., relative to Fe) are given in Table 2. The internal statistical mean error (usually derived from the line‐to‐line scatter) in the abundance ratios is given in column (3). The typical line‐to‐line abundance (i.e., relative to H) scatter for a given species is 0.08 dex, with a range from 0.04 dex (Si i) to 0.13 dex (Ni i). Given the uncertainties mentioned above in the S and Na results, we suggest uncertainties of 0.15 dex in their abundances. The Zr abundance uncertainty (0.10 dex) is assumed to be similar to the typical per line abundance scatter of other species. Columns (6)–(8) of Table 1 show the changes which result in the abundance ratios when altering each ofT_eff, log g, and ξ by the indicated amounts and holding the other parameters fixed.

The Li abundance was determined from the λ6708 Li i resonance feature by comparing the measured line strength with that of syntheses conducted using the linelist of King et al. (1997). We find log N(Li) = 2.53 ± 0.05, an abundance that is high relative to other warm stars of similar metallicity; therefore, we also examine the⁶Li/⁷Li ratio. First, a precise estimate of the wavelength shift to the rest frame was estimated from 14 Fe i lines having accurate wavelengths in the NIST on‐line database. Macroscopic broadening (parameterized with a Gaussian) was determined by synthesizing the profiles of eight different Fe i features using abundances from their equivalent widths, convolving these with Gaussians, and comparing these with the observed spectra. The left panel of Figure 3 shows a synthesis of the λ6708 Li i region containing no⁶Li and having log N(⁷Li) = 2.53 as determined from the measured line strength. The synthesis, assuming no⁶Li, is a wholly satisfactory fit to the data.

Zoom In Zoom Out Reset image size

The right panel of Figure 3 shows a synthesis with the same total Li, but ⁶Li/⁷Li = 0.15. To "improve" the quality of the fit, we have (a) shifted the data redward by an additional 4 mÅ with respect to our optimal value used in the left panel and (b) decreased our best‐estimate FWHM of the convolving Gaussian by 0.023 Å. Both of these changes are 2–3 σ excursions each as estimated from our multiple measures. Still, the synthesis appears to be slightly too weak (despite decreased macroscopic broadening) and the synthetic red core and wing too strong (despite the redward shift applied to the data). We believe that ⁶Li/⁷Li≲0.15 is a conservative limit to the⁶Li content of BD +00°2058A. Lower but nonzero values, though, are certainly allowed. A caveat is that the above analysis implicitly assumes that the photospheric velocity distribution (both its mean and width) is similar for Fe i and Li i. The rigorous accuracy of this usual assumption is unclear, as are the effects of, e.g., convective inhomogeneities.

Our result of 6130 ± 65 K for BD +00°2058A is in good accord with the infrared flux method estimate of 6065 ± 83 K by Alonso, Arribas, & Martinez‐Roger (1996) and somewhat warmer than the 6000 ± 42 K photometric estimate of Carney et al. (1994, hereafter C94); the agreement remains reasonable given the respective errors, and the higher Strömgren‐based reddening of E(B−V) = 0.03 [assuming E(b−y) = 0.72E(B−V)] of Schuster & Nissen (1989, hereafter SN89). A comparison of the dereddened Strömgren photometry of SN89 with the synthetic R. L. Kurucz (1992, private communication) grid colors in the (b−y)₀ versus c₀ plane indicates log g = 4.13 for [m/H] = -1.2 and log g = 4.19 for [m/H] = -1.0. These values are in excellent agreement with our spectroscopic estimate of log g = 4.19 ± 0.19. The gravity and near‐turnoff T_eff does not cleanly establish a dwarf or subgiant classification.

Our "metallicity" estimate [Fe/H] = -1.16 is a factor 2–4 larger than the moderate resolution Ca ii K–based estimate of Ryan & Norris (1991; −1.74), the high‐resolution, low S/N spectroscopic estimate (−1.51) of C94, and the Strömgren‐based value (−1.41) of SN89. The reason for these discrepancies is unclear. Table 2 indicates that our T_eff would have to be overestimated by 410 K to yield [Fe/H] = -1.5; this would result in unacceptable abundance‐excitation trends. Interestingly, our metallicity is in good agreement with the C94 estimate of [m/H] = -1.27 for the cooler proper motion companion BD +00°2058B; the Ryan & Norris (1991) [Fe/H] estimate for BD +00°2058A is also substantially lower than for the cooler component.

4.2.1. Age Estimate

Age, kinematics, and detailed abundances can be examined to address the Population classification of BD +00°2058A. First, we revisit the relatively low 10.5 Gyr age of SN89 by comparing our inferred T_eff and log g values with evolutionary models. Our [Fe/H] abundance, the scaling relation of Salaris, Chieffi, & Straniero (1993), and the assumption of Z_⊙ = 0.017 indicate that Z = 0.0018 isochrones are most appropriate for comparison with BD +00°2058A.

The left panel of Figure 4 plots BD +00°2058A in the log g versus T_eff plane with isochrones having Y = 0.23 and Z = 0.002 from Straniero & Chieffi (1991). BD +00°2058A appears to be right at the turnoff with an age of 14 Gyr, although an age as young as 10 Gyr is allowed. The right panel plots BD +00°2058A plotted with the Y = 0.23 and Z = 0.001 Yale96 isochrones;² this metallicity is appropriate for, e.g., lower metallicity determinations noted above. The star resides right at the turnoff and is very old (20 Gyr); about 4 of the 6 Gyr difference between the two sets of isochrones is because Yale96 isochrones are ∼100 K hotter at the metallicity, gravity, and rough age of our star. In sum, BD +00°2058A resides right at the turnoff and has an age similar to the oldest globulars, although an intermediate age is allowed.

Zoom In Zoom Out Reset image size

4.2.2. Distance and Kinematics

4.2.3. Abundance Signatures

Our Li abundance for BD +00°2058A, log N(Li) = 2.53, is significantly larger than the mean value (∼2.2) seen in other warm metal‐poor stars constituting the so‐called Spite plateau. Figure 7 of Lambert, Heath, & Edvardsson (1991) indicates that Li abundances this large are not seen until higher metallicities of [Fe/H]≳ - 0.5. The origin of the high Li abundance is relevant to ongoing debate about the Spite plateau. Is the large Li abundance of BD +00°2058 due to the extraordinary action of Galactic Li enrichment over and above a near‐primordial Spite plateau abundance, or has BD +00°2058 suffered less Li depletion than other Spite plateau stars from a higher primordial Li abundance?

One Li production mechanism is spallation reactions between cosmic rays and the ISM characterized by an isotopic ratio of ⁶Li/⁷Li∼1. For a factor of 2 Li enhancement above a primordial level (having no⁶Li); one thus expects a current isotopic ratio of ⁶Li/⁷Li∼0.25. Our conservative isotopic upper limit of ≲0.15 provides no evidence of spallation production of Li for this star. A second mechanism is the speculative ν‐process in supernovae (Woosley et al. 1990). Li production in this scenario should be accompanied by copious B production. Space‐based UV spectra of the λ2500 B i lines in BD +00°2058 could discern such B enrichment.

Another Galactic Li enrichment mechanism is the⁷Be transport mechanism (Cameron 1955), where⁷Be is produced in the deep interior of asymptotic branch (AGB) stars and quickly transported outward and converted to⁷Li via e⁻ capture. In principle, this⁷Li can enrich the ISM through mass loss, or enrich the surface of a companion via mass transfer. Observational evidence of this process comes from Li‐rich AGB stars in the Small Magellanic Cloud (Smith & Lambert 1989). An important observation is that Li‐rich metal‐poor AGB stars also demonstrate n‐capture overabundances and large heavy‐to‐light s‐process abundance ratios. Could Li‐rich material from a metal‐poor AGB star have contaminated the photospheric material (directly or prior to formation) of BD +00°2058? Our n‐capture abundances are unremarkable for this star, as is the heavy‐to‐light (e.g., [Ba/Zr] = -0.27) ratio. The n‐capture abundances thus provide no evidence of Li production carrying an s‐process signature. However, as kindly pointed out to the author by D. Lambert, such a signature may not be observable if the contaminating material is super Li rich [log N(Li)≳4.0]. However, the metal‐poor SMC AGB stars are all Li rich [log N(Li)∼3], but not super Li rich; contaminating Li abundances this low should result in observable s‐process enhancements not seen.

The metal‐rich J‐type AGB star WZ Cas does demonstrate a superrich Li abundance [log N(Li)∼6; Denn, Luck, & Lambert 1991] but no s‐process enhancement. If a similar mechanism of Li production with no neutron source also operates in a metal‐poor context, this could explain the Li abundance of BD +00°2058A. Another intriguing possibility is Li production in pre‐AGB first‐ascent giants. A number of field K giants are observed to be Li rich. That most of these objects are inferred to have circumstellar shells from far‐IR photometry has led de la Reza, Drake, & da Silva (1996) to propose this is a short‐lived but normal phase of evolution of low‐mass giants in which⁷Li is produced via the⁷Be transport process acting via "cool bottom processing" (Wasserburg, Boothroyd, & Sackmann 1995). If true, observational signatures of other abundance alterations in low‐mass giants (C→N, Ne→Na, Mg→Al) may be relevant.

Interestingly, such abundance signatures may exist in BD +00°2058A, although more observational work is needed to confirm these. First, our value of [C/Fe] = -0.09 ± 0.04 is some 0.3–0.4 dex below the metal‐poor C i–based abundances presented by Tomkin et al. (1995) and Reynolds et al. (1992). An uncertainty in a claimed C deficiency is the poorly understood ∼0.3 dex systematic difference between C i– and CH‐based C abundances in metal‐poor stars (e.g., Wheeler, Sneden, & Truran 1989); the latter are typically in agreement with our value for BD +00°2058A. Second, our estimate of [Na/Fe] = +0.21 ± 0.15 is larger than values derived in some studies of metal‐poor stars. Unfortunately, homogeneous analyses of numerous stars with metallicities higher and lower than our value for BD +00°2058A are lacking, and comparison of extant work suggests study‐to‐study differences and/or significant intrinsic scatter in metal‐poor Na abundances, which complicates interpretation of our result. Additional complications are that the line strength of our weak Na feature is uncertain, and the other feature has a very large solar equivalent width. A determination from the λλ5682, 5688 lines would be welcome.

Finally, we noted that our [Mg/Fe] ratio may be ∼0.15 dex below similarly metal‐poor stars; however, the presence of intrinsic scatter in metal‐poor [Mg/Fe], and possible study‐to‐study differences cloud the reality of this modest difference. Regardless, the C, Na, and Mg results seem to form a consistent picture of material having undergone mixing on the red giant branch. If true, then contamination of the BD +00°2058A photosphere with Li‐rich material produced in an evolved star having undergone cool bottom processing would be a promising means to explain the Li abundance of BD +00°2058A. Of course, the "action" of both Li enrichment and in situ mixing/Li destruction is also possible.

Reanalysis of our data motivated by discussion with J. Laird concerning his unpublished analyses of BD +00°2058A reveal that the detailed accuracy of our adopted gf‐values may be an important source of systematic error in our derived parameters and abundances. As an example, we repeated the parameter analysis using the critical compilation of Fe i gf‐values from Lambert et al. (1996, hereafter LHLD). The mean difference in the log gf values for the 22 Fe i lines in common (in the sense LHLD, this work) is −0.011 ± 0.013 dex (s.d. of 0.058 dex).

This excellent on average agreement seems satisfactory, but the detailed dependences on excitation potential are such that, for the lines in common, employing the LHLD gf‐values instead of our values yields estimated T_eff values some 110 K lower than values from our gf‐values. As a result, the spectroscopic log g estimate would be 0.25 dex lower. Most of the difference seems to be attributable to differences in gf‐values for moderate excitation lines (∼2.5 eV). Such a difference between the OB91 and BKK gf‐values (used here) and the Oxford results (heavily relied upon by LHLD) has been noted before. As emphasized by K98, subsets of lines or the detailed dependence(s) of gf differences can significantly influence derived parameters and abundances of metal‐poor stars despite overall agreement between different sources of gf‐values.

Since it is unclear which, if either, choice of oscillator strengths is correct, we have simply preferred a "sole‐source vendor" (or nearly so) of Fe i gf‐values here. How would the present results differ if, e.g., LHLD's critical gf‐values were employed? The derived [Fe/H] would be −1.23, which is in closer agreement with C94's estimate for BD +00°2058B (−1.27) but still discrepant with previous estimates for BD +00°2058A itself. The Li abundance would be lowered by 0.1 dex, and its discrepancy becomes more marginal—especially considering its (now) subgiant status. As can be judged from Table 2, the [C/Fe] and [Na/Fe] ratios do not change appreciably, and thus would remain odd. The other abundance ratios change by only a few hundredths of a dex and would remain unremarkable.

Given the revised parameters, the isochrones in the left and right panels of Figure 4 suggest ages of 11 and 15–16 Gyr, and the star would be evolved just past the turnoff. A revised luminosity of M_V = 3.58 is found, and this gives an estimated distance of 203 pc. The resulting kinematics are U = +130, V = -123, and W = -307 km s⁻¹. The large W and V_sp (355 km s⁻¹) are indicative of a halo classification. These kinematics are much less remarkable in the V‐[m/H] distributions of C96's Figure 5.

It is with both pleasure and sadness that the author remembers the capable professional assistance of the late Mike Marcario, McDonald Observatory telescope operator, in obtaining the observations presented here.