HUBBLE SPACE TELESCOPE NEAR-ULTRAVIOLET SPECTROSCOPY OF BRIGHT CEMP-s STARS^*

STIS (Kimble et al. 1998; Woodgate et al. 1998) observations of HD 196944 and HD 201626 were obtained as part of Program GO-12554, using the E230M echelle grating, centered at 2707 Å, and the NUV Multianode Microchannel Array (MAMA) detector. There was one observational sequence of three individual exposures for each star, taken on 2012 April 29 (HD 201626) and 2012 September 22 (HD 196944). The total integration time was about 7.9 ks per star. The 006 × 02 slit yields a resolving power of R ∼ 30,000. Our setup produced a wavelength coverage from 2280 Å–3070 Å in a single exposure. The observations were reduced and calibrated using the standard calstis pipeline. The S/N of the combined spectrum varies from ∼45 pix⁻¹ near 2300 Å, to ∼70 pix⁻¹ near 2700 Å, to >90 pix⁻¹ near 3070 Å.

The NUV spectra of the program stars around the Mg ii doublet at 2800 Å are shown in Figure 1. For comparison, HST/STIS spectra of the CEMP-no star BD+44°493 (obtained during this same program and analysed by Placco et al. 2014a) is also shown. The effective temperatures and surface gravities of the program stars are comparable to BD+44°493, but their metallicities are higher by 1.4−2.3 dex. To better illustrate the intrinsic differences between these CEMP-s stars and CEMP-no stars, Figure 2 shows portions of the NUV spectra around the lines of four neutron-capture elements for the same three stars shown in Figure 1. Note in particular the absence of absorption by Cd i, Os ii, Lu ii, and Pb i for BD+44°493, which signal a lack of neutron-capture enhancement in BD+44°493 relative to HD 196944 and HD 201626.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The Keck/HIRES (Vogt et al. 1994) spectra of both stars were gathered over two observing runs: 2004 October 1 (program ID: C23H and PI: I. Ivans) and 2007 June 5 and 7 (program ID: U100Hb and PI: M. Bolte). HD 196944 was observed in three 300 s integrations in both runs. HD 201626 was observed with three 450 s integrations in program C23H and three 240 s integrations in program U100Hb. The U100Hb data cover the the wavelength range 3725 Å $\lesssim \;\lambda \;\lesssim$ 7990 Å, with resolution $R\equiv \lambda /{\rm{\Delta }}\lambda \;=$ 103,000. The C23H data cover bluer wavelengths (2990 Å $\lesssim \;\lambda \;\lesssim$ 5850 Å), with resolution R = 48,000.

The C23H data were reduced using standard tasks in IRAF,¹⁶ including bias subtraction, bad-pixel interpolation, and wavelength calibration; and in FIGARO,¹⁷ including flat-fielding, light cosmic ray excision, both sky and scattered-light subtraction, and extraction of the one-dimensional spectra. The same reduction steps were performed on the U100Hb data using the MAKEE pipeline¹⁸ (2008 version 5.2.4). Final processing of the HIRES data, including contiuum normalization and co-addition of the trios of extracted spectra were done using SPECTREdsg (D. S. Gregersen 2015, private commnication), an updated version of the SPECTRE package (Fitzpatrick & Sneden 1987).

We determined self-consistent stellar model-atmosphere parameters using spectroscopic constraints on the abundances of neutral and ionized species of the same element. For CEMP and other cool stars, equivalent-width measurements (EWs) of clean unblended iron (Fe) lines are usually employed. However, the NUV region of our stars is extremely rich and complex. Thus, we derived our atmospheric parameters using the EWs of Fe i and Fe ii lines in the optical wavelength range. We then employed this model atmosphere to derive the remaining abundances at both NUV and optical wavelengths.

Our EWs were obtained using the following tools:

• (i)  
  Fitting Gaussian profiles to the observed atomic lines, using the Robospect package (Waters & Hollek 2013).
• (ii)  
  Fitting Gaussian and Voigt profiles, using the SPECTREdsg package, an updated version of the SPECTRE package (Fitzpatrick & Sneden 1987).
• (iii)  
  Fitting Gaussian and Voigt profiles, using the THIMBLES package (T. R. Anderton 2015, in preparation).

We compared the measurements obtained using the THIMBLES and SPECTREdsg packages. For 100 lines in common between two sets of independent measurements, the difference in reduced equivalent widths, $\mathrm{REW}\equiv \mathrm{log}(\mathrm{EW})/\lambda =0.01\pm 0.05.$

The Robospect line lists were based on the compilation of Roederer et al. (2012), and data retrieved from the VALD database (Kupka et al. 1999) and the National Institute of Standards and Technology Atomic Spectra Database (NIST; Kramida et al. 2013). The SPECTREdsg and THIMBLES line lists are based on those used in Ivans et al. (2003, 2006), supplemented and updated by the work of Blackwell-Whitehead & Bergemann (2007), Nilsson & Ivarsson (2008), Meléndez & Barbuy (2009), Simmerer et al. (2013), Wood et al. (2013, 2014a, 2014b), Lawler et al. (2014), Ruffoni et al. (2014), and Andersson et al. (2015).

The atmospheric parameters of the stellar models we employed in our analysis were derived from these HIRES spectra, employing the following spectroscopic constraints. The effective temperature (${T}_{\mathrm{eff}}$) was determined by minimizing any trend between the abundances of Fe i absorption lines with the excitation potential (χ); the surface gravity (log g) was determined by the equilibrium balance between the abundances of Fe ii and Fe ii; and the microturbulent velocity (ξ) was determined by minimizing the trend between the Fe i abundances derived for the individual lines and their REWs.

For HD 196944, we derived the best stellar model-atmosphere parameters to be 5170 ± 100/1.60 ± 0.25/−2.41 ± 0.25/1.55 ± 0.10 [${T}_{\mathrm{eff}}$ (K)/log g (cgs)/[Fe/H]/ ξ (km s⁻¹)]. Our values are consistent with previous determinations from the literature, listed in Table 1. For HD 201626, we derived the best stellar model-atmosphere parameters to be 5175 ± 150/2.80 ± 0.45/−1.51 ± 0.25/1.30 ± 0.10. Despite the large literature on HD 201626, the derived stellar model-atmosphere parameters are not in very good agreement between several different authors, as seen in Table 1. The Karinkuzhi & Goswami (2014) model-atmosphere parameter estimates agree reasonably well with ours (within 50 K for ${T}_{\mathrm{eff}}$ and 0.1 dex for [Fe/H]), apart from a 0.55 dex difference in log g. Both studies derived ${T}_{\mathrm{eff}}$ values from spectroscopic constraints on the abundances. However, Aoki & Tsuji (1997) adopted a ${T}_{\mathrm{eff}}$ based on the infrared flux method (Blackwell et al. 1980), and Sneden et al. (2014) adopted one based on the $V-K$ color of the star (Ramirez & Melendez 2005), albeit with a low reddening value ($E(V-K)\;\simeq$ 0.171). Table 1 also contains additional colour-based estimates we have made of some of the stellar parameters.

Table 1.  Atmospheric Parameters

${T}_{\mathrm{eff}}$	log g	[Fe/H]	ξ	Method	Reference
(K)	(cgs)		(km s−1)
HD 196944 (Literature)
5170	1.60	−2.41	1.55	a	This work
5250	1.70	−2.45	1.90	a	Zacs et al. (1998)
5250	1.80	−2.25	1.70	a	Aoki et al. (2002)
⋯	⋯	−2.40	⋯	a	Van Eck et al. (2003)
5170	1.80	−2.46	1.70	a	Roederer et al. (2008)
5310	1.75	−2.39	1.65	a	Roederer et al. (2014a)
HD 201626 (Literature)
5175	2.80	−1.51	1.30	a	This work
5190	2.25	−1.30	2.30	a	Vanture (1992)
4800	2.00	−1.50	2.00	b	Aoki & Tsuji (1997)
5120	2.25	−1.40	1.02	a	Karinkuzhi & Goswami (2014)
4800	1.30	−1.90	2.00	c	Sneden et al. (2014)
4550	⋯	⋯	⋯	d	Casagrande et al. (2011)
4852	⋯	⋯	⋯	e	McDonald et al. (2012)
HD 201626 (Additional Color-based Estimates)
4561a	⋯	⋯	⋯	c	B −V from Wenger et al. (2000)
4593a	⋯	⋯	⋯	c	ubvy from Hauck & Mermilliod (1997)
4971a	⋯	⋯	⋯	c	V from Wenger et al. (2000); K from 2MASS
⋯	⋯	−1.29a	⋯	f	ubvy from Hauck & Mermilliod (1997)
⋯	⋯	−1.59a	⋯	g	ubvy from Hauck & Mermilliod (1997)
⋯	⋯	−1.23a	⋯	f, h	ubvy from Hauck & Mermilliod (1997)
⋯	⋯	−1.53a	⋯	g, h	ubvy from Hauck & Mermilliod (1997)

Notes.

(a) Spectroscopic constraints, (b) IRFM (Blackwell et al. 1980), (c) color-Teff (Ramirez & Melendez 2005), (d) color-Teff (Casagrande et al. 2011), (e) SED modelling, (f) uvby-metallicity (Calamida et al. 2007); empirical relation, (g) uvby-metallicity (Calamida et al. 2007); theoretical relation, (h) converted to Kraft & Ivans (2003, 2004, p. 33) metallicity scale.

^aEstimates made by adopting the $E(B-V)$ value based on the dust maps of Schlafly & Finkbeiner (2011) at the IRSA Galactic Dust Reddening and Extinction Service (http://irsa.ipac.caltech.edu/applications/DUST/). However, as noted by Sneden et al. (2014), because the star is only some 300 pc away, one would expect to see a smaller dereddening effect from a fraction of the total $E(B-V)$ value (and thus, an even cooler photometric color-${T}_{\mathrm{eff}}$).

Download table as:  ASCIITypeset image

To investigate the temperature estimate for HD 201626 issue further, we performed the following experiment. Using a stellar-atmosphere model with the Aoki & Tsuji (1997) parameter estimates, we find log (Fe i) = 5.50 and log (Fe ii) = 5.60; with the Sneden et al. (2014) parameters, we find log (Fe i) = 5.57 and log (Fe ii) = 5.30. Thus, demanding a cooler ${T}_{\mathrm{eff}}$ for this star drives the derived metallicity lower, much closer to the [Fe/H] adopted by Sneden et al. (2014). If there was an independent method of establishing the metallicity of the star other than by the spectroscopic methods we have employed in this study, a lower [Fe/H] might indicate that we have the ${T}_{\mathrm{eff}}$ wrong for HD 201626. However, the abundances derived from these other models do not satisfy the spectroscopic constraints that we employed.

While we acknowledge the tension between the spectroscopically derived ${T}_{\mathrm{eff}}$ values and those derived over a broader wavelength range for HD 201626, the underlying cause is beyond the scope of this analysis. Our stellar model-atmosphere parameters are derived from superb high-resolution spectra; our line list is extensive (see Table 2); and, as we show in Section 3, we find excellent agreement between neutral and ionized states of multiple elements. Of greatest importance for this analysis is that our spectroscopically derived stellar model-atmosphere parameters produce spectrum-synthesis calculations that match the observed data in the NUV wavelengths.

Table 2.  Derived Optical Abundance Estimates

Ion	λ	χ	$\mathrm{log}\;{gf}$	$\mathrm{log}\epsilon$ (X)	$\mathrm{log}\epsilon$ (X)
	(Å)	(eV)		HD 196944	HD 201626
Equivalent Width Analysis
C i	4771.75	7.49	−1.87	⋯	8.56
C i	5380.34	7.68	−1.62	⋯	8.26
C i	6587.61	8.54	−1.00	⋯	8.11
Na i	5682.63	2.10	−0.70	<4.10	⋯
Na i	5688.19	2.10	−1.40	3.90	4.50
Na i	6160.75	2.10	−1.25	⋯	5.01
Mg i	4571.10	0.00	−5.57	5.34	6.39
Mg i	4702.99	4.34	−0.44	5.61	⋯
Mg i	5172.69	2.71	−0.45	5.74	⋯
Mg i	5183.60	2.72	−0.24	5.68	⋯

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

Figure 3 shows a portion of the NUV spectra of both HD 196944 and HD 201626, in a region with a number of Fe i and Fe ii features. The dots represent the observed spectra, the solid line is the best abundance fit, generated with the atmospheric parameters determined from the optical spectra. The dotted and dashed lines limit the shaded area, which encompass a 1.0 dex difference in [Fe/H]. As can be seen from inspection of this figure, the fit is excellent.

Zoom In Zoom Out Reset image size

Heliocentric radial velocities for the program stars were determined from clean unblended lines in the optical spectra. For HD 196944, we derived ${V}_{r}=-176.40\pm 0.60$ km s⁻¹ on MJD 53279.255346; ${V}_{r}=-169.40\pm 0.30$ km s⁻¹ on MJD 54256.557629; and ${V}_{r}=-174.30\pm 0.20$ km s⁻¹ on MJD 54574.636680.¹⁹ This star has a number of V_r measurements from the literature: ${V}_{r}=-174.76\pm 0.36$ km s⁻¹ (Aoki et al. 2002); ${V}_{r}=-174.10\pm 0.40$ km s⁻¹ (Van Eck et al. 2003); ${V}_{r}=-169.29\pm 0.08$ km s⁻¹ (Lucatello et al. 2005); ${V}_{r}\;=-168.49\pm 0.11$ km s⁻¹ (Lucatello et al. 2005); ${V}_{r}\;=-166.40\pm 0.30$ km s⁻¹ (Roederer et al. 2008); and ${V}_{r}\;=-166.80\pm 0.70$ km s⁻¹ (Roederer et al. 2014a). These radial-velocity variations indicate that HD 196944 is a member of a binary or multiple system. However, there is no published value of its period based on these measurements. From the values determined by this work and from the literature, we were able to estimate a period for HD 196944. We have used a program described in Buchhave et al. (2010), based on the formalism of Pál (2009), which models a Keplerian orbit to the data. Assuming initial guesses of the period and ephemeris, we were able to determine a period of 1325 ± 12 days, and an eccentricity of 0.015 ± 0.036.

For HD 201626, we derived ${V}_{r}=-150.20\pm 1.70\;\mathrm{km}\;{{\rm{s}}}^{-1}$ on MJD 53279.325761; and ${V}_{r}=-140.00\pm 0.30\;\mathrm{km}\;{{\rm{s}}}^{-1}$ on MJD 54258.560998. Literature values include 26 measurements from McClure & Woodsworth (1990), and also the following: ${V}_{r}=-149.40\pm 0.80\;\mathrm{km}\;{{\rm{s}}}^{-1}$ (Van Eck et al. 2003); ${V}_{r}=-145.70\pm 0.70\;\mathrm{km}\;{{\rm{s}}}^{-1}$ (Nordström et al. 2004); ${V}_{r}\;=-141.60\pm 1.20\;\mathrm{km}\;{{\rm{s}}}^{-1}$ (Karinkuzhi & Goswami 2014). HD 201626 is a confirmed binary with a period of 1465 ± 15 days, and an eccentricity of 0.103 ± 0.038 (McClure & Woodsworth 1990). From our derived radial-velocities and the values with available MJD, we were able to determine a period of 1470 ± 9 days, and an eccentricity of 0.090 ± 0.051, which are consistent with the values from the literature.

Carbon, nitrogen, and oxygen abundances were determined for both stars, using the NUV and optical spectra, from both EW analyses and spectral synthesis calculations.

Carbon abundances were determined from optical CH features at 4246 Å and 4313 Å. For HD 196944, both regions are consistent with log (C) = 7.10 and, for HD 201626, log (C) = 8.40. The lower panels of Figures 4 and 5 show the synthesis for the 4246 Å region in both stars. We derived an isotopic ratio ¹²C/¹³C = 4 for HD 196944, and ¹²C/¹³C = 50 for HD 201626. The high ¹²C/¹³C for HD 201626 is consistent with AGB models, and the low value for HD 196944 can be explained by its evolutionary stage, in which internal mixing slightly decreased the ¹²C/¹³C ratio. We also attempted to determine the carbon abundance from the C i atomic features at 2478 Å and 2967 Å. For HD 196944, the log (C)=7.10 value determined for the optical spectra can be well-reproduced for the 2478 Å line (upper panel of Figure 6). The agreement between abundances determined from the atomic (NUV) and molecular (optical) features is worth noting, since previous works (Asplund 2005; Collet et al. 2007) have suggested that 3D effects on the CH and C₂ bands could lead to overestimates of [C/Fe] of $+0.5$ to $+0.8$ dex in metal-poor subgiants. For HD 201626, both regions are heavily blended with other species, and it was not possible to determine abundances. We also calculated the carbon-abundance corrections for both stars, based on the procedure described by Placco et al. (2014b). This procedure takes into account the evolutionary status of the star (from the log g value) to find a correction for the carbon depletion, which occurs during stellar evolution on the giant branch. Since HD 196944 is a red horizontal-branch star, we take its carbon correction only as a lower limit. Using the [N/Fe] = 0.0 case, we find corrections of > +0.23 dex for HD 196944, and 0.02 dex for HD 201626.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Nitrogen abundances were determined from the NH feature at 3360 Å. The line list was provided by Kurucz (1993), following the procedure described in Aoki et al. (2006). For HD 196944, the best value is log (N) = 6.50 (upper panel of Figure 4) and, for HD 201626, log (N) = 6.70 (upper panel of Figure 5).

Oxygen abundances were determined from NUV spectral synthesis of OH features for HD 196944, and spectrum synthesis of the [O i] 6301 Å line for HD 201626. For HD 201626, we were also able to measure the O triplet near 7700 Å. However, we do not include the abundances from these lines in our analysis, for reasons we describe below.

The O-triplet features possess high excitation potentials, and are expected to suffer from non-LTE effects. For HD 201626, LTE analysis of the O-triplet EWs yields log (O) $=\;8.15,$ with a standard deviation of 0.17 dex (based on the line-to-line scatter). Applying the recommended correction of Takeda (2003) for each of the three lines, we derive log (O)$\;=\;8.05,$ which is closer to that of the abundance we derive from spectrum-synthesis calculations of the [O i] 6301 Å feature. A star with atmospheric parameters close to those of HD 201626 was studied by Garcia Perez et al. (2006) (HD 274949: 5090/2.76/−1.51), who independently calculated non-LTE abundance corrections for the triplet. Their Δlog (O) correction is $-0.10$ dex, in agreement with the value we find adopting the recommended correction of Takeda (2003). However, in a comparison against Takeda (2003), Fabbian et al. (2009) find the necessary non-LTE corrections to be significantly larger than those reported by Takeda (2003) for stars warmer than 5500 K. Fabbian et al. (2009) also note that the stellar metallicity plays an important role in the degree of the non-LTE effect. For the stars included in the study of Ramirez et al. (2013) that are closest in their stellar parameters to HD 201626 (HIP 60719: 5250/2.7/−2.42 and HIP 18235: 5010/3.2/−0.73) the corrections to the LTE log (O) abundance derived from the triplet are Δlog (O) of +0.66 dex and −0.29 dex, an order of magnitude difference. Thus, in our analysis of HD 201626, we rely on the abundance derived from the [O i] 6301 Å line.

There are several OH features in the 2965–2975 Å region of HD 196944 (lower panels of Figure 6). These were all be reasonably fit with a log (O) = 6.60. However, we conservatively adopt a 0.3 dex uncertainty, due to the presence of a few unknown features in this region.

Abundances for Ti, Cr, and Ni from the NUV spectra were determined for HD 196944 with an EW analysis. No such measurements were made for HD 201626, since it is more metal-rich and the same NUV lines are heavily blended. We derived Mn ii abundances from spectral synthesis of NUV features at 2933 Å and 2949 Å for both stars, because these lines are broadened by hyperfine splitting of the ⁵⁵Mn isotope. Figure 7 (upper panels) shows the synthesis of the 2933 Å Mn ii line in the NUV spectrum for both program stars. We were also able to determine the abundance of Ge in both stars. This element belongs to the transition between iron-group and neutron-capture elements. Three lines (2651 Å, 2691 Å, and 3039 Å) were measured for HD 196944, and one line (2691 Å) for HD 201626. The lower panels of Figure 7 show the spectral synthesis for the Ge i 2691 Å line.

Zoom In Zoom Out Reset image size

A number of iron-peak element abundances were determined from EW analysis of the optical spectra of HD 196944 and HD 201626. Comparing the optical and NUV determinations, we find good agreement for Ti, Cr, and Ni for HD 196944, and Mn for both stars. For elements with two different ionization states measured in HD 196944, abundances are within 0.04 dex for Ti and 0.22 dex for Cr obtained from the optical data, and 0.02 dex for Ti, 0.18 dex for Cr, and 0.40 dex for Ni from the NUV data. The good agreement between the Ti i and Ti ii also confirms our spectroscopically determined log g values. The large difference for Ni lines in the NUV is mainly due to the difficulty of measuring EWs for the Ni ii lines.

For the neutron-capture elements (Cu to Pb), all of the abundances listed in Tables 4 and 5 were determined from spectrum-synthesis calculations, using both optical and NUV spectra. For the optical spectra, some elements had their abundances first estimated from their EW that were used as starting points for the synthesis.

There are a number of unblended Zr ii lines in the NUV region of the spectra of these stars. For both HD 196944 and HD 201626, the agreement between the individual abundances is generally good (between 0.1–0.2 dex), apart from the 2699 Å feature, which is consistently lower by about 0.5 dex. Ba ii lines are broadened by hyperfine splitting (hfs) of the ¹³⁵Ba and ¹³⁷Ba isotopes; so even though we also determined Ba abundances from EW analysis, the final average values come from spectral synthesis only. Figure 8 shows the optical spectral synthesis of the Ba ii 5853 Å line for both stars, the Pb i 4057 Å line for HD 196944, and the Nd ii 5319 Å line for HD 201626. The EW Ba abundances for HD 196944 agree within 0.1 dex with the spectral synthesis for the 4554 Å, 5853 Å, and 6141 Å lines (log (Ba) = 1.0). For HD 201626, we report the abundance from the spectrum synthesis of the 5853 Å feature. The synthesis of the 6141 Å and 6406 Å features are consistent with this value, but the 4554 Å region is too complex to model well.²¹

Zoom In Zoom Out Reset image size

Figure 9 shows the spectral synthesis of Lu, Hf, Au, and Nb for HD 196944, and Pb (2833 Å line) for both stars. Pb abundances determined from neutral species are more affected by non-LTE effects than abundances determined from ionized species. For the stellar parameters of the stars studied in this work, this effect can increase the Pb abundance by up to 0.5 dex (Mashonkina et al. 2012). This effect is somewhat counterbalanced by considering 3D modeling (Siqueira Mello et al. 2013). For simplicity, we consider the combination of both effects to be within the uncertainties of the spectral synthesis. Overall, we were able to reproduce the observed spectra with our line list, even with difficult blends, such as the one for the Au i 2675 Å line. For HD 196944, there is a small discrepancy between the Pb abundances determined from the 2833 Å (log (Pb) = 1.35) and 4057 Å (log (Pb) = 1.50) lines.

Zoom In Zoom Out Reset image size

Figure 10 shows the optical and NUV abundances for HD 196944, compared with AGB yields from the model presented in Placco et al. (2013), with M = 1.3 M${}_{\odot }$ and [Fe/H] = −2.5, as well as three additional models from Bisterzo et al. (2010), with [Fe/H] = −2.1 and M = 0.9, 2.0, and 4.0 M${}_{\odot }.$ Although all the models well-reproduce the second peak of the s-process (Ba–Nd), the more massive AGB stars over-produce the first-peak s-process elements (Sr, Y, and Zr). In addition, Pb is under-produced by the [Fe/H] = $-2.1$ models, and the observed values are closer to the [Fe/H] = $-2.5$ model. Bisterzo et al. (2011) also studied HD 196944, and found the best match with a model of M = 1.5 M${}_{\odot },$ with no initial enhancements of r-process elements.

Zoom In Zoom Out Reset image size

The Bisterzo et al. (2010) models also calculate the evolution of s-process element abundances with [Fe/H], which is a useful diagnostics for selecting AGB models with different masses and ¹³C-pocket efficiencies. Figure 11 shows the [Pb/Fe], [Ba/Fe], and [Pb/Ba] values for HD 196944 and HD 201626 compared with four of the model prescriptions from Bisterzo et al. (2010). The [Pb/Ba] values for HD 196944 and HD 201626 are consistent with the s-process nucleosynthesis in low-mass low-metallicity AGB stars with the highest neutron exposures possible.

Zoom In Zoom Out Reset image size

Bisterzo et al. (2010) also point to the importance of measuring Nb abundances in CEMP stars, to distinguish between the intrinsic (star on the thermally pulsing AGB or post-AGB phase) and extrinsic (main-sequence or red giant-branch star—CEMP-s stars). As noted by Wallerstein & Dominy (1988), an intrinsic AGB star is expected to be technetium-rich (Tc; Z = 43), ⁹³Zr-rich, and ⁹³Nb-poor, with [Zr/Nb] ∼ 1. The s-process abundances of an extrinsic AGB star are the result of pollution from a former AGB star onto the atmosphere of what once was the lower mass star in a binary system where the abundance of Nb (which has a single stable isotope) is a result of β-decay from ⁹³Zr, leading to [Zr/Nb] ∼ 0 (see e.g., the CEMP-$r/s$ star CS 29497−030, with [Zr/Nb] = −0.27, Ivans et al. 2005). For HD 196944, for which these two elemental abundances were derived from the NUV spectra, we find [Zr/Nb] = $+0.47\pm 0.21.$ However, we note that our Nb abundance is derived from the weak Nb ii feature at 2950 Å.

We also investigated the properties of HD 196944 and HD 201626 by comparing their derived elemental abundances with the grid of binary-evolution models of Abate et al. (2015b). This grid consists of about 290,000 binary stars distributed in the ${M}_{1}-{M}_{2}-{\mathrm{log}}_{10}a-{M}_{\mathrm{PMZ}}$ parameter space, where ${M}_{\mathrm{1,2}}$ are the initial masses of the primary and secondary stars, respectively, a is the initial orbital separation of the system, and ${M}_{\mathrm{PMZ}}$ is the mass of the partial-mixing zone. The mass of the partial-mixing zone is a parameter of the nucleosynthesis model that relates to the amount of free neutrons available for neutron-capture processes in AGB stars.²² In this study we used model set B of Abate et al. (2015b), which allows efficient angular-momentum loss during mass transfer and large wind mass-accretion efficiencies over a wide range of orbital separations.

The best-fitting model to the observed abundances of each star is determined following Abate et al. (2015b). The model ages are selected between 10 and 13.7 Gyr, bracketing the expected ages of very metal-poor halo stars. The orbital periods of the models are chosen to reproduce the observed periods within the spatial resolution of our grid (${\rm{\Delta }}{\mathrm{log}}_{10}(a/{R}_{\odot })=0.1,$ which corresponds to ${\rm{\Delta }}{\mathrm{log}}_{10}({P}_{\mathrm{orb}}/\mathrm{days})=0.15$). To constrain the evolutionary stage, we select model stars that reproduce the measured surface gravities within the observational uncertainties. For the model stars that fulfill these criteria, we determine the model that minimizes ${\chi }^{2},$ calculated by taking into account the abundances of the light elements up to atomic number 12 (Mg), and all neutron-capture elements from Ga (Z = 31) to Pb (Z = 82). The elements with atomic numbers between 13 (Al) and 30 (Zn) are not considered in the calculation of ${\chi }^{2},$ because they are not in general produced by low-mass stars ($M\lt 3{M}_{\odot },$ see, e.g., Karakas et al. 2009; Cristallo et al. 2011). In cases where only upper limits are available for a given element, an observational uncertainty of 0.3 dex is adopted to compute ${\chi }^{2},$ if the model abundance value is higher than the observed upper limit. This value was chosen because it is the largest observational uncertainty from our abundance determinations.

The initial abundances used in the models (up to Ge) are predicted by Kobayashi et al. (2011) in their models of Galactic chemical evolution at [Fe/H] = −2.3. For elements heavier than Ge, we use Solar System scaled values. It must be kept in mind that the impact of the choice of the initial set of abundances is negligible when studying CEMP-s stars. This is because the abundances of most key elements (e.g., C, N, Na, Mg, and all neutron-capture elements) have large variations during AGB evolution. The fit is made by combining the abundances of C and N. This is because, in giants, some amount of C is converted to N at the bottom of the convective envelope during dredge up. Even though the exact amount is very uncertain, the total C+N is conserved, therefore the total C+N predicted by the models is supposed to reproduce the observed abundance. For this comparison, we thus do not employ the carbon-abundance corrections, just the observed values.

Figures 12 and 13 show the best-fitting models for both HD 196944 (${M}_{1,i}=0.9{M}_{\odot },$ ${M}_{2,i}=0.86{M}_{\odot },$ for [Fe/H] = $-2.2$), and HD 201626 (${M}_{1,i}=0.9{M}_{\odot },$ ${M}_{2,i}=0.76{M}_{\odot },$ for [Fe/H] = $-2.2;$ ${M}_{1,i}=1.6{M}_{\odot },$ ${M}_{2,i}=0.59{M}_{\odot },$ for [Fe/H] = $-1.5$) as solid lines. The abundances from this work are shown as filled dots with error bars. The shaded areas encompass the minimum and maximum abundances predicted by models that: (i) have an age between 10 and 13.7 Gyr; (ii) reproduce the observed period within the grid resolution and log gwithin the observational uncertainties; and (iii) have a final ${\chi }^{2}$ that is less than three times the ${\chi }^{2}$ of the best fit. For HD 201626, we ran two models, with metallicities Z = 10⁻⁴ and $5\times {10}^{-4},$ which correspond to [Fe/H] = −2.2 and −1.5, respectively, with the solar abundance set of Asplund et al. (2009). Also shown in the plot are the initial parameters of the best models ([Fe/H], M₁, M₂, ${M}_{\mathrm{PMZ}}$) and the final orbital period (given as input). Comments on the model-matching for each star are given below.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

4.2.1. HD 196944

4.2.2. HD 201626