THE KAPTEYN MOVING GROUP IS NOT TIDAL DEBRIS FROM ω CENTAURI^*

We target stars from the Kapteyn group, an apparently coherent kinematic structure within the solar neighborhood that has recently been linked, both kinematically and chemically, to the GC ω Centauri (WdB10). Eggen (1996) listed 33 stars kinematically associated to the Kapteyn group (earlier defined by Eggen 1977), from which there are 7 that are variable. Our sample collects 14 of the remaining 26 non-variable stars, all of them included in the study of WdB10, which contains a total of 16 stars, extracted from the study of Eggen (1996).

We also include in our sample stars from the "ω Cen group" of Meza et al. (2005), identified as having specific angular momenta similar to that of the ω Centauri cluster and slightly retrograde orbits (−50 km s⁻¹ < V_rot < 0 km s⁻¹). These stars are part of the catalogs of metal-poor field stars of Beers et al. (2000) and Gratton et al. (2003).

A comparison sample of field stars was also observed. These stars were selected from the catalog of Gratton et al. (2003), such that they span the range of stellar parameters of stars in the Kapteyn and ω Cen groups.

Table 1 lists our target stars with their fundamental atmospheric parameters, as determined below.

Table 1.  Spectroscopic Stellar Parameters of Our Target Stars

Star	Vhel	Teff	$\mathrm{log}g$	[Fe/H]
	(km s−1)	(K)	(dex)	(dex)
Kapteyn Group Stars
CD-30 1121	105.25 ± 0.33	4955 ± 65	2.07 ± 0.16	−1.919 ± 0.06
CD-62 1346	125.74 ± 0.35	5524 ± 83	2.16 ± 0.18	−1.399 ± 0.06
HD 110621	222.17 ± 0.33	6212 ± 97	4.04 ± 0.20	−1.522 ± 0.07
HD 111721a	21.53 ± 0.35	4990	2.60	−1.380
HD 13979	52.38 ± 0.36	5023 ± 145	1.47 ± 0.22	−2.441 ± 0.08
HD 181007	−1.72 ± 0.35	4979 ± 51	2.23 ± 0.13	−1.721 ± 0.05
HD 181743	29.50 ± 0.37	5998 ± 123	4.26 ± 0.26	−1.795 ± 0.09
HD 186478	31.45 ± 0.35	4568 ± 85	0.88 ± 0.19	−2.500 ± 0.07
HD 188031	−144.21 ± 0.40	6288 ± 81	4.24 ± 0.16	−1.649 ± 0.06
HD 193242	−127.85 ± 0.39	5031 ± 54	2.35 ± 0.14	−1.778 ± 0.05
HD 208069	−166.59 ± 0.41	5065 ± 46	2.47 ± 0.12	−1.741 ± 0.04
HD 21022	122.85 ± 0.33	4524 ± 62	0.89 ± 0.15	−2.204 ± 0.05
HD 215601	−36.65 ± 0.36	4972 ± 45	1.67 ± 0.12	−1.460 ± 0.04
HD 215801	−90.87 ± 0.40	6065 ± 184	3.69 ± 0.18	−2.183 ± 0.06
ω Cen Group Stars
BD+02 3375	−397.77 ± 0.49	6009 ± 160	4.01 ± 0.18	−2.190 ± 0.06
CD-61 0282	220.35 ± 0.34	5815 ± 70	4.30 ± 0.16	−1.234 ± 0.06
HD 113083a	227.91 ± 0.46	5800	4.35	−0.930
HD 121004	245.63 ± 0.33	5651 ± 48	4.31 ± 0.11	−0.784 ± 0.04
HD 140283	−170.36 ± 0.39	5788 ± 79	3.56 ± 0.13	−2.393 ± 0.05
HD 148816	−47.62 ± 0.34	5876 ± 43	4.16 ± 0.11	−0.790 ± 0.04
HD 193901	−171.39 ± 0.39	5721 ± 44	4.42 ± 0.11	−1.109 ± 0.04
HD 194598	−247.04 ± 0.40	5998 ± 47	4.33 ± 0.12	−1.148 ± 0.04
HD 3567	−47.60 ± 0.35	6130 ± 51	4.06 ± 0.12	−1.200 ± 0.04
HD 84937	−15.12 ± 0.39	6280 ± 150	4.00 ± 0.20	−2.089 ± 0.07
Field Stars
HD 102200	161.26 ± 0.33	6113 ± 62	4.15 ± 0.15	−1.222 ± 0.05
HD 116064	140.26 ± 0.38	6010 ± 195	4.16 ± 0.22	−1.848 ± 0.07
HD 128279	−75.44 ± 0.35	5245 ± 97	2.79 ± 0.23	−2.123 ± 0.08
HD 134439	310.04 ± 0.34	4956 ± 69	4.60 ± 0.17	−1.440 ± 0.06
HD 134440	310.77 ± 0.35	4795 ± 66	4.60 ± 0.16	−1.426 ± 0.06
HD 142948	29.28 ± 0.35	5024 ± 87	2.52 ± 0.14	−0.720 ± 0.05
HD 145417a	8.68 ± 0.37	4900	4.60	−1.349
HD 151559	16.67 ± 0.37	5167 ± 59	2.54 ± 0.13	−0.769 ± 0.05
HD 17072	62.43 ± 0.36	5398 ± 58	2.37 ± 0.14	−1.128 ± 0.05
HD 184266	−348.82 ± 0.47	5908 ± 80	2.20 ± 0.18	−1.463 ± 0.06
HD 190287	143.74 ± 0.33	5108 ± 47	2.78 ± 0.12	−1.429 ± 0.04
HD 199289	−5.95 ± 0.34	5842 ± 54	4.27 ± 0.13	−1.090 ± 0.05
HD 211998	32.10 ± 0.34	5233 ± 46	3.25 ± 0.12	−1.519 ± 0.04
HD 219617	13.28 ± 0.35	5981 ± 61	4.40 ± 0.15	−1.427 ± 0.05
HD 221580	−17.35 ± 0.36	5245 ± 60	2.30 ± 0.15	−1.215 ± 0.05
HD 222434	16.88 ± 0.36	4517 ± 53	1.21 ± 0.13	−1.770 ± 0.05
HD 222925	−38.64 ± 0.36	5710 ± 60	2.32 ± 0.14	−1.371 ± 0.05
HD 23798	88.88 ± 0.34	4396 ± 60	0.64 ± 0.15	−2.220 ± 0.05
HD 26169	−34.42 ± 0.35	5010 ± 68	2.06 ± 0.16	−2.396 ± 0.06
HD 83212	109.21 ± 0.32	4543 ± 57	1.17 ± 0.13	−1.504 ± 0.05
HD 9051	−71.99 ± 0.38	4884 ± 49	2.08 ± 0.12	−1.652 ± 0.04

Note.

^aReference star (see Section 3 and Table 3).

Download table as:  ASCIITypeset image

Optical spectroscopic observations of our sample stars were carried out using the MIKE spectrograph on the 6.5 m Magellan/Clay Telescope at Las Campanas Observatory. We employed the standard setup with the narrowest slit (width 0.35 arcsec), leading to spectral resolutions of $R\simeq 83,000$ and 65,000 in the blue and red CCDs, respectively. The full wavelength range covered by the standard setup of the MIKE spectrograph is 320–1000 nm. The S/N of our spectra varies from star to star, but it is typically in the range from 150 to 200 per pixel at 6000 Å.

The MIKE spectra were reduced (trimmed, corrected for overscan, flat-fielded, and corrected for background/scattered light) with the CarnegiePython pipeline, which also provides a wavelength mapping on the data with the help of ThAr exposures taken throughout each observing night. These reduced spectra were further processed for continuum normalization, merging, and co-adding of multiple exposures (when needed) using common IRAF⁹ tasks within the echelle package.

The continuum normalization procedure can introduce errors in elemental abundance measurements. To estimate this effect, we tested different normalization parameters such as the order of the polynomials used to fit the upper envelopes of the data and the sigma-clipping values. On a representative spectrum from our dataset, we estimated changes in our EW measurements of up to 5%, which translate to about 0.02 dex in [X/H]. These errors are smaller than our estimated total error of about 0.05 dex, which is dominated by the model uncertainties. Moreover, the changes in EW values were calculated relative to our preferred normalization method. If all EWs are measured consistently, i.e., on spectra that have been normalized using the same parameters, the errors are minimized because we employ strict differential analysis using stars with similar stellar parameters, for which the normalization procedure traces very similar continua. Even if the continuum errors were of order 0.02 dex, they would not be large enough to invalidate our main chemical tagging conclusions. As can be inferred from Figures 3 to 7, we would have had to under- or overestimate abundance ratios of some of our stars by up to 0.5 dex to make a number of our stars be consistent with the ω Centauri patterns.

2.2.1. Radial Velocities (RVs)

High-resolution, high-S/N near-infrared spectra for the target stars were collected using the CRIRES (Kaeufl et al. 2004) mounted on the VLT UT1 telescope. The observations were carried out in Service Mode (Program 090.B-0605(B), 29 hr, P.I. J.C.) from 2011 October through 2012 March.

The CRIRES setting employed a slit width of 04 with the grating order #52, yielding a spectral resolution of $R\simeq 50,000$. This configuration allows us to observe He i 10830 Å near the center of chip 3, which covers a range not affected by vignetting.¹⁰ Under this setup, detector 3 has a wavelength reference of 10827 Å (wavelength at the middle of the detector) and a wavelength coverage from 10800 to 10860 Å.

Data reduction was performed using the CRIRES pipeline (Common Pipeline Library recipes, version 2.4), running under the GASGANO environment.¹¹ Master dark subtraction, bad pixel map, flat-fielding, and corrections for nonlinearity effects were applied. Wavelength calibration for the spectra was applied using ThAr wavelamp frames. In order to improve the wavelength solution a high-resolution ISAAC spectrum¹² with telluric lines was used as reference and a conversion equation was determined from the difference between the nominal wavelength of the telluric lines and the obtained positions from the hot, fast rotating star HR 1996. On average, the telluric lines appeared redshifted by ∼0.5 Å with respect to their nominal wavelength.

Since the ThAr catalog lines used during the wavelength calibration of the spectra are in vacuum wavelengths, the He i line does not appear at 10830 Å (air wavelength), but instead it is found slightly redshifted, at 10833 Å. In order to properly join the information from the Arcturus stellar lines atlas by Hinkle et al. (1995, in vacuum wavelengths) and the atmospheric water vapor lines listed by Breckinridge & Hall (1973, in air wavelengths), the IAU standard equation of Edlén (1953) was used to go from one to the other. The coefficients for the infrared regime were derived by Peck & Reeder (1972). Table 2 shows the air and vacuum wavelengths for the stellar and telluric lines expected to be found in the CRIRES spectra.

Telluric lines were identified based on the spectrum of the hot, fast rotating star HR 1996. The wavelength-calibrated spectrum of this star is shown in Figure 12 of the appendix, where the most prominent telluric lines (in the vacuum wavelength rest frame) are marked.

Figures 13–16 show all our CRIRES rest-frame spectra in the region of interest. For all the spectra, the normalization of the continuum was performed using the continuum task of IRAF. Our science targets are indicated by the top blue spectrum on each panel, while the green spectrum, shown for comparison purposes, corresponds to the telluric standard star shifted in wavelength to match the telluric features in the science target. The left and right dashed lines marked the positions of the Si i and He i lines, respectively.

The fundamental atmospheric parameters of our stars (${T}_{\mathrm{eff}},\mathrm{log}g,[\mathrm{Fe}/{\rm{H}}]$) were determined using the conditions of relative excitation and ionization equilibrium for iron lines (Fe i and Fe ii). The EWs of as many as possible of the iron lines employed by Ramírez et al. (2013) were measured by fitting Gaussian profiles to the observed spectral lines within IRAF's task splot.

For each star, we started with a set of guess parameters (obtained from previous calculations found in the literature) and computed iron abundances using the curve-of-growth method on a line-by-line basis. Kurucz odfnew model atmospheres and the spectrum synthesis code MOOG were employed for these calculations. We then examined the correlations between relative iron abundance (with respect to a specific reference star, as explained below) and excitation potential as well as reduced EW ($\mathrm{REW}=\mathrm{logEW}/\lambda$). We also examined the difference between mean iron abundance inferred from Fe i and Fe ii lines separately. The initial set of parameters was then modified iteratively until the correlations disappeared and the Fe i and Fe ii lines provided the same mean iron abundance. On average, 98 lines (80 Fe i and 18 Fe ii) were used to derive the iron abundances. The minimum number of Fe transitions measured was 21 Fe i and 8 Fe ii for HD 215801.

When using the excitation/ionization method described above, it is customary to use the Sun as the standard star. However, our sample is heterogeneous regarding spectral type and it also is biased toward very low metallicity. Employing the Sun as reference is not ideal due to the propagation of systematic uncertainties, which are very difficult to control in highly heterogeneous samples like ours. Thus, instead we adopted three different reference stars that are representative of the three main groups of targets that we have available: dwarf and subgiant stars, cool dwarf stars, and RGB stars. The three reference stars and their adopted stellar parameters are listed in Table 3.

For each of the reference stars we performed a relative analysis using the Sun as reference and adopting an absolute solar iron abundance of ${A}_{\mathrm{Fe}}^{\odot }=7.50$. Thus, while systematics still exist, they can be better controlled by investigating the parameters of the reference stars. The relative parameters within each group of stars are expected to be precise and internally consistent. In any case, the stars chosen as reference are all well-studied bright halo stars, and the parameters we derived using the Sun as reference are in reasonably good agreement with other previous and independent estimates. Conservatively, we expect these parameters to be uncertain at the level of 100 K in ${T}_{\mathrm{eff}}$, 0.1 in $\mathrm{log}g$, and 0.05 dex in [Fe/H].

We note that the reference stars are on the metal-rich side of the stellar parameter distribution of their respective samples. Strictly speaking, it would be better to employ reference objects with stellar parameters that are centered on the distributions of each sample. However, good reference stars at low metallicities are significantly less common than metal-rich ones and, while not ideal, our procedure is certainly superior to one where the Sun, a very metal-rich star in this context, is employed as the only reference star.

Our derived stellar parameters are listed in Table 1. The errors were computed by propagating the errors in the slopes of the iron abundance versus excitation potential and reduced EW added in quadrature with the errors due to the uncertainty in the mean iron abundances from Fe i and Fe ii lines measured separately. For [Fe/H] we also added in quadrature the standard error of the line-to-line dispersion.

Figure 1 shows our metallicity measurements for the stars that have also been listed by WdB10 and the PASTEL catalog (Soubiran et al. 2010). The dashed line in the top panel corresponds to the 1:1 line, where practically all the stars with measurements from the PASTEL catalog lie. In the case of the stars with metallicity measurements in the study of WdB10, the differences are higher (∼0.3 dex). All the metallicities measured by those authors are higher than the ones measured in this work and in the PASTEL catalog.

Zoom In Zoom Out Reset image size

The differences in metallicity between the measurements of WdB10 and the values derived in this work and the PASTEL catalog are propagated into the stellar parameters. For example, the stellar parameters of the star HD 181007 derived by WdB10 and the ones listed in the PASTEL catalog are in disagreement by more than 250 K for ${T}_{\mathrm{eff}}$ and 1.0 and 0.5 dex for $\mathrm{log}g$ and [Fe/H], respectively. More worrying cases include HD 186478 and HD 181743. For the former, WdB10 report [Fe/H] = −1.92 while four studies included in PASTEL reveal a value closer to −2.6 dex. In the case of HD 181743 the [Fe/H] and $\mathrm{log}g$ values given by WdB10 are 1.1 and 0.2 dex lower, respectively, than most previously reported measurements, while the ${T}_{\mathrm{eff}}$ is in good agreement with the values reported by the PASTEL catalog.

Moreover, when comparing our [Fe/H] values with those from the PASTEL catalog, we find an average difference of 0.01 ± 0.09 dex. Our iron abundances have a typical error of the order of 0.05 dex, whereas a heterogeneous catalog like PASTEL is expected to have [Fe/H] errors in the range 0.05–0.10 dex. Thus, the predicted error in the average difference quoted above should be between 0.07 and 0.11 dex, in good agreement with the observed value of 0.09 dex. In stark contrast, when the WdB10 data are compared to the PASTEL catalog, the average difference in [Fe/H] is 0.31 ± 0.26 dex. The large dispersion seen in this comparison, and the small one seen in our case, strongly suggest that the [Fe/H] values from WdB10 are unreliable.

We searched for as many as possible of the spectral lines listed in Table 1 of Ramírez et al. (2011) and Table 4 of Ramírez et al. (2014) in order to measure chemical abundances of elements other than iron. Since that list was constructed for stars of solar metallicity, many of those lines are too weak in the spectra of our target stars. Thus, the line list was expanded with spectral lines listed in the studies by Nissen et al. (2002), Reddy et al. (2003), Neves et al. (2009), and Bensby et al. (2013). The first of these references provides a thorough and reliable list of lines for very metal-poor stars in general and the other three references, while made for studies of somewhat metal-rich stellar populations, include spectral lines that are also useful in the analysis of stars with effective temperatures and surface gravities significantly different from the solar values. On average, the number of transitions employed to derived the chemical abundances was three lines of each of the elements considered (Na i, Mg i, Al i, Ba ii, and O i). For Ca i the number of lines employed was, on average, ten per star. The individual EW measurements are listed in Table 4.

Table 4.  List of Atomic Transitions and Equivalent Width Measurements for Our Target Stars

			Equivalent Width
Wavelength	Species	χ	$\mathrm{log}{gf}$	CD-30 1121	CD-62 1346	HD 110621	(Cont.)
(Å)		(eV)		(mÅ)	(mÅ)	(mÅ)
7771.94	8.0	9.15	2.25	12.3	66.8	43.7	⋯
7774.16	8.0	9.15	1.67	8.5	51.8	34.1	⋯
7775.39	8.0	9.15	1.00	6.0	37.7	27.7	⋯
4751.82	11.0	2.10	−2.08	...	...	...	⋯
4982.82	11.0	2.10	−1.00	...	...	...	⋯
5682.64	11.0	2.10	−0.77	6.4	13.0	6.9	⋯
5688.21	11.0	2.10	−0.48	13.1	25.1	10.1	⋯
6154.23	11.0	2.10	−1.55	...	...	...	⋯
6160.75	11.0	2.10	−1.25	...	6.8	...	⋯
4730.04	12.0	4.34	−2.39	8.6	24.3	7.2	⋯
5711.09	12.0	4.34	−1.73	27.8	41.9	19.7	⋯
6318.72	12.0	5.10	−1.95	5.1	11.3	...	⋯

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

With the stellar parameters listed in Table 1, we used the abfind driver of MOOG and Kurucz odfnew model atmospheres to derive elemental abundances for our sample stars. As for the stellar parameters, the absolute abundances of the reference stars were first computed using the Sun as reference. Then, we measured only relative abundances for the rest of our targets, using the appropriate reference star for each group. Table 5 presents the metallicity and the derived chemical abundances [X/Fe] for our target stars. The number of lines on which the quoted errors are based is indicated in parenthesis.

The sodium lines in the spectrum of HD 84937 were all found to have very low local S/N and in most cases the line was not even detected. Thus, no Na abundance is given for this star.

Abundances of Al are given only for a limited number of stars due to the difficulty in measuring the EWs of the very weak Al lines available in our optical spectra. In fact, in most of the cases where no Al abundance is reported in Table 5, the Al lines were not even detected.

Given the importance of the oxygen abundances for this work, and the fact that they were derived using the 777 nm O i triplet, which is known to be severely affected by departures from the LTE approximation, we made every effort to obtain as reliable as possible [O/Fe] values. We employed the grid of non-LTE corrections by Ramírez et al. (2007) to derive line-by-line non-LTE oxygen abundances of all stars more iron-rich than [Fe/H] = −1.6. This grid has a lower limit in metallicity of [Fe/H] = −1.4, but it has been shown to be reliable in extrapolations down to [Fe/H] = −1.6 (see, for example, Ramírez et al. 2011, 2013). The non-LTE corrected oxygen abundances of our three reference stars were computed differentially with respect to the Sun. Then, relative oxygen abundances were calculated for the other [Fe/H] > −1.6 stars differentially, each with respect to its corresponding reference star. Note that all our three reference stars have [Fe/H] > −1.4.

A significant fraction of our stars are more iron-poor than [Fe/H] = −1.6. For these objects, we employed LTE abundances to calculate oxygen abundances relative to their corresponding reference stars. However, for the reference stars, we adopted the non-LTE oxygen abundances derived as described above. This mixing of LTE and non-LTE oxygen abundances will certainly introduce a systematic error, but it will be somewhat mitigated by the fact that the calculations are done differentially for each star relative to another object of comparable atmospheric parameters. These "pseudo" non-LTE corrected oxygen abundances were adopted for the [Fe/H] < −1.6 stars.

To test the reliability of our adopted oxygen abundances, we computed non-LTE corrections for all our stars using the grid by Fabbian et al. (2009), which includes the full [Fe/H] range covered by our stars. Nevertheless, we also had to extrapolate this grid in order to apply it to our data. This is because nearly all the LTE abundances we derived are higher than the upper limit given in that grid. In any case, we noticed that the [O/Fe] versus [Fe/H] relation that results using the grid by Fabbian et al. is similar in shape to that obtained using our adopted oxygen abundances, even at low metallicity. In fact, the most notable difference is a roughly constant offset of about +0.15 dex in [O/Fe] due to the fact that the oxygen abundances calculated using the grid by Fabbian et al. are systematically higher.

In a number of reliable elemental abundance studies, some of which employ oxygen features other than the 777 nm triplet (e.g., Asplund & García Pérez 2001; Sneden & Primas 2001; Meléndez & Barbuy 2002; Meléndez et al. 2006), the [O/Fe] abundance ratios of very metal-poor stars are seen to lie in a plateau at a level of about +0.5, similar to what we obtained with our adopted oxygen abundances. The [O/Fe] values inferred using the grid of Fabbian et al., on the other hand, would place the [O/Fe] plateau at a significantly higher level ($\simeq +0.65$). For this reason, we prefer to adopt the oxygen abundances as described before, and not those computed with the Fabbian et al. grid. The comparison, however, shows that in a relative sense, the two methods provide similar results. This means that stars that are too oxygen-poor or too oxygen-rich relative to the bulk of halo stars could be equally identified using either set of [O/Fe] results.

The 777 nm oxygen triplet lines were not detected in the spectra of HD 134439, HD 134440, and HD 13979. Therefore, no oxygen abundances are reported for these three stars.

Later in this paper, we use the elemental abundances derived from the optical spectra to perform a chemical tagging experiment, i.e., to attempt to associate our samples to known stellar populations for which previous chemical composition studies exist. Most of these previous works have adopted a pure LTE approach or used the oxygen forbidden lines, which are formed under LTE conditions, to derive [O/H]. Therefore, for our chemical tagging experiment, the impact of non-LTE effects on the abundances is not as important as being able to perform a fair comparison instead.

The EW of the He i 10830 Å line in our program stars was measured with the IRAF's task splot. We use the deblend option when a telluric line appears contaminating the wings or the continuum around the line, which occurs in six of our program stars (four stars of the Kapteyn/ω Cen group sample and two of the field sample). For another star of the field comparison sample, HD 211998, a strong telluric line is at exactly the same position as the He line, preventing us from saying anything about it.

We also measured, when possible, the EW of the line directly using the Gaussian fit of splot, as a validation test for our measured values. However, when telluric lines are near the He i line, the position of the wings is not easy to identify and the line sometimes could be asymmetric, leading to a less accurate Gaussian fit. Figure 2 shows the fit using both procedures for the He i line of CD-621 346. From the spectrum, the wings of the He i line are difficult to identify and the direct Gaussian fit gives a lower value than the deblending option, which takes into account the fact that part of the wings of one line and its neighbor can overlap.

Zoom In Zoom Out Reset image size

The EW measurements derived using the deblending task, EW₁, and the direct Gaussian fit, EW₂, are listed in Table 6, where we also list the S/N achieved by our CRIRES observations at the location of the He i line. Errors in EW were calculated using the formula of Cayrel (1988), which uses the FWHM of the line, the S/N of the spectrum, and the dispersion of the instrument (5.2 × 10⁻³ nm pix⁻¹ for the setup used in our CRIRES observations). A minimum value of 10 mÅ as lower limit is used to separate stars that show detectable helium in their spectra from those that do not. A similar cut-off was found by Dupree et al. (2011) for the spectra of the stars where the helium line is not apparent.

In most of the spectra, there is a good agreement between the two EW values measured as mentioned before. The same behavior was found by Smith et al. (2014), comparing the EWs of the He i line in giants of M13. Nonetheless, in some cases the difference is non-negligible and not explained by the listed errors, mainly because the line is in blends, and the Gaussian fit cannot take into account the part of the spectrum that is missed in the blended wing. Because of that, we adopt hereafter the EWs measured using the deblending option.

Although the He i 10830 Å line has not been measured frequently in a systematic way, there are nine stars from our sample that have published EWs, which are listed in Table 7 together with our derived EW values from CRIRES spectra. A relatively good agreement is found for five stars; HD 194598 is the only star that has two measurements in the literature and both values agree with our derived value. However, there are some stars that have important differences, which might be due to an intrinsic variability of the line depending on the chromospheric activity. This is further discussed in Section 4.2.

One of the main distinguishing chemical patterns of GC stars is an anticorrelated behavior of O and Na abundances, with a fraction of the stars showing O–Na levels similar to field stars, but another important fraction showing depleted O and enhanced Na (Carretta et al. 2009a; Johnson & Pilachowski 2010; Gratton et al. 2012, and references therein). The latter group of stars has been associated to a second stellar generation that formed out of material enriched by the hotter branches of hydrogen burning that took place in stars of the first generation (Ventura et al. 2011; Gratton et al. 2004). Therefore, if the Galactic field contains a fraction of stars originating from GCs, one would expect to find some of these O-poor, Na-rich second-generation stars as well (e.g., Altmann et al. 2005; Ramírez et al. 2012). Applying this to the context of the present work, therefore, if the Kapteyn group were tidal debris from ω Centauri, then an unambiguous indication that this is the case would be the identification among the group's stars of some with this particular O–Na pattern.

Figure 3 shows the O and Na abundances for our target stars from the Kapteyn group (red triangles), ω Cen group (blue diamonds) and field stars (green circles). The O and Na abundances of GC stars (open gray circles from Carretta et al. 2009a) and giant stars from ω Centauri (open magenta squares from Marino et al. 2011) are included. The two star symbols correspond to the two GC second-generation stars in the field found by Ramírez et al. (2012).

Zoom In Zoom Out Reset image size

The oxygen abundances of the stars from both the Kapteyn and ω Cen groups are indistinguishable from those of the stars of the field sample, which are located at the base of the GC Na–O anticorrelation. HD 151559 appears as the only star from our whole sample that could be a second-generation star from a GC in the field. This star has the lowest O abundance and the third most enhanced Na abundance of our full sample.

ω Centauri is known to harbor at least three different populations, with mean metallicities at ∼ −1.7, −1.2, and −0.8 dex (Johnson & Pilachowski 2010). These populations are consistent with the multiple MSs found through its color–magnitude diagram, namely the red, blue, and anomalous MS (as defined by Bedin et al. 2004; Bellini et al. 2010). Spectroscopic analyses of RGB stars from these three different populations by Marino et al. (2011) and Gratton et al. (2011) reveal that the metal-poor and metal-intermediate populations present a marked O–Na anticorrelation and at least two different generations of stars (primordial and intermediate, i.e., the first and second generations). It is important to note that this conclusion does not depend on the specific bins adopted to separate the three metallicity populations, thus discarding the possibility that the O–Na anticorrelation seen in both populations is due to the overlapping of the populations. Only the metal-rich population (at [Fe/H] ∼ −0.8 dex) is practically all composed of second-generation stars and, contrary to the others, it actually shows a slight O–Na correlation (Gratton et al. 2011). In particular, Gratton et al. (2011) show that there is an extreme component of second-generation GC stars (as defined by Carretta et al. 2009a) that is consistent with being O-poor, Na-rich, and also He-rich (up to $Y\simeq 0.387$ dex) compared to stars of an intermediate component that are also O-poor and Na-rich, but with almost the same He abundance as the primordial generation. This first generation is mostly indistinguishable from the halo field stars with the same metallicity. Similar to monometallic GCs, ω Centauri has a prominent fraction of second-generation stars (Johnson & Pilachowski 2010). According to the criteria to separate the primordial, intermediate, and extreme components in GCs developed by Carretta et al. (2009a) and the O and Na abundances reported by Johnson & Pilachowski (2010), the fractions of stars belonging to the primordial, intermediate, and extreme components in ω Centauri are 42%, 44%, and 14%, respectively, in agreement with the measured fractions in monometallic GCs (Johnson & Pilachowski 2010). Nonetheless, ω Centauri has a prominent fraction of extreme component stars in its metal-rich population, even reaching 81% of that population, the greatest fraction found in any GC so far (Johnson & Pilachowski 2010).

Motivated by this, our target stars were divided into three different metallicity ranges, in order to better visualize whether second-generation stars are present among the Kapteyn or the ω Cen groups. Figure 4 shows the O–Na plane for the three different metallicity regimes, i.e., metal-poor, [Fe/H] < −1.65; metal-intermediate, $-1.65\leqslant [\mathrm{Fe}/{\rm{H}}]\lt -1.00$; and metal-rich, $[\mathrm{Fe}/{\rm{H}}]\geqslant -1.00$. In each panel, ω Centauri's RGB stars from the catalog of Johnson & Pilachowski (2010) are also plotted, according to the different metallicity bins.

Zoom In Zoom Out Reset image size

From the metallicity distribution of the Kapteyn group stars, most of them (9 out 14 stars) are metal-poor, while five are in the metal-intermediate range and there are no stars in the metal-rich group. Figure 4 shows there is no evidence of an O–Na anticorrelation in the two groups. The same result is found for the stars in the ω Cen group, which are indistinguishable from the Galactic halo stars in each metallicity bin.

GC stars also follow a Mg–Al anticorrelation (Carretta et al. 2009a), which is the result of high-temperature H-burning chains. In contrast, disk and halo stars of the MW are mostly confined to nearly solar Al and Mg abundances (Reddy et al. 2003, 2006). Therefore, if stars in the Kapteyn and ω Cen groups showed that kind of behavior, it would be suggestive of a GC origin.

Aluminum lines are weak or undetectable for most of our target stars. Figure 5 shows the stars for which we could measure an abundance for this species—only three from the Kapteyn group and five from the ω Cen one—preventing us from tracing any possible Mg–Al pattern. All these stars, nevertheless, have similar abundances to those found for the field star samples.

Zoom In Zoom Out Reset image size

Could it be that our sample of Kapteyn group stars just so happens to contain only first-generation stars stripped from ω Centauri? As in other GCs, the second-generation stars in ω Centauri are more centrally concentrated than the first-generation (see Bellini et al. 2009, and references therein). Therefore, it can be expected that a larger fraction of the stripped stars from the cluster belong to the first generation. However, as we explain immediately below, if all our Kapteyn and ω Cen group stars originated in ω Centauri, it is unlikely that all of them could be at the same time from the first stellar generation of the cluster.

Considering that the fraction of second-generation GC stars in the Galactic halo is of the order of 2%–3% (Martell et al. 2011; Ramírez et al. 2012), the models of Vesperini et al. (2010) allow us to estimate that at most 50% of the Galactic halo stars come from GCs. Nonetheless, if we consider that our stars are typical Galactic halo stars, from our observations, we have one out of 45 ($\simeq 2.2\%$) stars with Na and O abundances consistent with GC second-generation stars, compatible with the 2%–3% fraction expected.

Based on the concentrations derived by Gratton et al. (2011), the number of first- and second-generation stars from the different populations could be estimated out to 10 arcmin, corresponding to about a fourth of ω Centauri's tidal radius. According to Table 1 and Figure 9 of their paper, the mix of stars at 10 arcmin from the center should be composed of 35%, 60%, and 5% from the metal-poor, metal-intermediate, and metal-rich populations, respectively. Within the metal-poor and metal-intermediate populations, ∼30% and 40% of the stars should correspond to second-generation stars. Therefore, assuming the stripped regions of ω Centauri contained a similar mix of populations, if the Kapteyn and ω Cen groups were part of ω Centauri's tidal debris, they should contain approximately 30% of second-generation stars: four and three stars in the Kapteyn and ω Cen groups, respectively. Although crude, this fraction of stars from the second-generation that could be stripped from ω Centauri is at odds with our our derived chemical abundances, which are consistent with only first-generation stars.

Figure 6 shows the EW of the He i line for all the stars in our program as a function of metallicity. ω Centauri's red giants observed by Dupree et al. (2011) are also plotted as open magenta squares, with the starred symbols indicating those stars with enhanced Al and Na abundances. The dotted line at 70 mÅ marks the typical maximum value found in non-active metal-poor stars from the Galactic field (Takeda & Takada-Hidahi 2011; Smith et al. 2012). All the stars in the ω Cen group have EW below 70 mÅ, making the group indistinguishable from the sample of comparison field stars. Stars of the Kapteyn group also appear with typical EWs, except CD-62 1346 which has a somewhat large EW of 91.2 mÅ. There is no correlation between the EW of the He line and the metallicity, as seen for the metal-intermediate group of stars in ω Centauri (Dupree et al. 2011).

Zoom In Zoom Out Reset image size

The ω Centauri giants with starred symbols in Figure 6 are second-generation star candidates based on their Na and Al abundances (Dupree et al. 2011). These authors found a better correlation between the detection of the chromospheric He i line and the enhanced Na–Al abundances than as a function of the metallicity. For stars in the Kapteyn and ω Cen groups, the measured He i EWs are generally smaller than those of the second-generation stars of Dupree et al. (2011) with seemingly enhanced helium. As discussed earlier, they also have O, Na, Mg, and Al abundances normal among field stars, which would be consistent with their He i EWs, assuming the latter to be indicative of helium content. For red giants in Messier 13, a GC with no significant helium enhancement measured (ΔY ∼ 0.04), Smith et al. (2014) found several He i EW values lower than 40 mÅ, similar to the distribution found for the Kapteyn and ω Cen groups.

Among the control sample of field stars, HD 222925 and HD 184266 strongly deviate from all the stars in all of our samples. HD 184266 has an EW of the order of magnitude of those found in active cool stars (see Sanz-Forcada & Dupree 2008), but no previous studies reported a measurement for the helium line for this star. HD 222925 has a He i EW about 5–6 times larger than the average value for the remaining stars in our field sample and, moreover, it has an enhanced [Ba/Fe] abundance (Figure 7) that makes it an intriguing target in its own right (see Section 4.5).

Zoom In Zoom Out Reset image size

Given that the He i 10830 Å line is associated with activity in cool stars, we also explore its variability among a few of the stars in our samples using our own data obtained for another program, as well as some measurements available in the literature. CRIRES spectra were obtained for HD 13979, HD 21022, and HD 215601 about a year after the observations reported in the present work. Using the same reduction process, the spectrum around 10830 Å was extracted, and in Figure 8 we show the normalized spectra at the two epochs for the three stars. Both HD 13979 and HD 215601 show an indistinguishable helium line shape between the two observations. In contrast, the He i line in HD 21022 changes from being not detected on the first run (EW smaller than 10 mÅ) to appearing in emission a year later, with an EW of 50 mÅ.

Finally, we found differences in the He i EW measurements in the literature for two stars in our ω Cen group (see Table 7). For HD 3567, Smith et al. (2012) determine an EW that is about half of that derived here, while for HD 140283, Takeda & Takada-Hidahi (2011) report an EW five times larger than ours. For HD 83212, from our field sample, our data show He i non-detections, while observations by Dupree et al. (2009) show the line in emission, with an EW of 43.6 mÅ.

In summary, out of twelve stars in our full sample with multiple measurements of the He i 10830 Å line, we find that four present significant variability between separate epochs. This suggests that the line should at least be used with caution as an abundance indicator.

Barium is one of the neutron-capture elements. It is the product of the s-process that takes place during the thermal pulses of AGB stars. Therefore, the barium abundance is a proxy for measuring the s-process contribution in a stellar system. ω Centauri has a unique [Ba/Fe]–[Fe/H] distribution that does not follow the expected distribution of MW or dwarf spheroidal (dSph) galaxy stars (e.g., Geisler et al. 2007; Marino et al. 2011). For the most metal-rich population (between $-1.5\;\leqslant$ [Fe/H] $\leqslant \;-0.5$ dex), ω Centauri's stars have an extreme overabundance of barium, [Ba/Fe] ∼ 0.7 dex, while the metal-poor stars experience a sharp increase in [Ba/Fe] with [Fe/H] between −2.0 and −1.5 dex. This is the result of the high contribution of AGB stars to the chemical enrichment process that occurs between the formation of the metal-poor and metal-intermediate components of the cluster (Marino et al. 2011). To date, only Majewski et al. (2012) and Casey et al. (2014a) seem to have identified stars that appear to be consistent with an origin in ω Centauri based on this unique Ba abundance pattern.

Figure 7 shows the [Ba/Fe]–[Fe/H] distribution for our target stars, in comparison with MW and ω Centauri stars from previous studies. MW stars predominantly have Ba abundances near the solar value, typical of the Galactic halo population. As a whole, neither Kapteyn nor ω Cen group stars follow a trend like the one found for ω Centauri stars, which puts into question any possible association between them.

However, there are two stars with a marked overabundance of Ba within our samples, compared to the other target stars with the same metallicity. CD-62 1346, a target from the Kapteyn group, has the most extreme Ba abundance of our sample, [Ba/Fe] ∼ 1.6 dex (WdB10 find [Ba/Fe] = 0.95 for this star, also the largest value for their sample). [Ba/Fe] determinations for our Kapteyn group sample range from −0.5 to 0.3 dex with an average value of [Ba/Fe] = −0.07, if CD-62 1346 is excluded. The [Ba/Fe] abundances derived by WdB10 have an average value of 0.21 dex, and 0.16 dex when excluding CD-62 1346, both higher than the determinations in the present work. However, the trend with metallicity even in WdB10 is still inconsistent with that of ω Centauri.

The other star with an overabundance of Ba is HD 222925, but it belongs to our field control sample and we discuss its case further in Section 4.5.

ω Centauri is thought to be the remnant nucleus of a completely disrupted dwarf galaxy (Dinescu et al. 1999; Bekki & Freeman 2003; Bekki & Norris 2006). Based on that, it seems necessary to address the possibility that the Kapteyn and ω Cen groups originated in the main body of the defunct progenitor galaxy instead of the GC itself.

Carretta et al. (2010a) compared detailed chemical abundances for ω Centauri RGB stars and the stars in the Sagittarius dSph galaxy and its nuclear GC, M54. Based on [Fe/H] metallicities, the authors concluded that ω Centauri's stars resemble the metallicity distribution found for M54 and Sagittarius dSph together. In particular, the metal-rich population of ω Centauri is consistent with the extent of the metallicity distribution of Sagittarius itself. Moreover, the α-element abundance for the surrounding nucleus of Sagittarius is similar to what is found for the metal-rich stars in ω Centauri while M54 shows [α/Fe] abundances in agreement with the metal-poor and metal-intermediate populations of ω Centauri.

The abundances of three alpha-elements were obtained from our optical spectra (O, Mg, and Ca), so we can compare the α-element abundances of our target stars with those measured for stars in dSph galaxies. Since most of the α-abundances in the literature are computed from those of Mg and Ca, we use these elements in our own determinations, leaving O aside.

Figure 9 shows the [α/Fe] abundance versus metallicity for our target stars, MW stars, and stars from the Sagittarius, Sculptor, and Carina dSph galaxies. The metallicity distributions of both the Kapteyn and ω Cen groups do not overlap with that of Sagittarius. As the GC ω Centauri appears consistent with the metal-rich stars ($[\mathrm{Fe}/{\rm{H}}]\geqslant -1.46$ dex) in Sagittarius, it appears straightforward to discard the idea that the Kapteyn and ω Cen groups' stars have an origin in the defunct host galaxy of ω Centauri, if such an extinct host was comparable to what we see today in Sagittarius.

Zoom In Zoom Out Reset image size

For Carina and Sculptor, and although they cover the same metallicity range as our target stars, the range of [α/Fe] variation for stars in those dSphs is very broad and incompatible with our stars. In fact, Kapteyn and ω Cen groups' stars follow, at least qualitatively, the same trend as MW field halo stars. At $[\mathrm{Fe}/{\rm{H}}]\leqslant -1.5$ dex, halo stars (Venn et al. 2004) have an average [α/Fe] abundance of 0.30 with a dispersion of 0.12 dex. The comparison field stars in our sample have a very similar average [α/Fe] abundance, of 0.30 dex with σ = 0.10 dex. For Kapteyn group stars, the same average is 0.33 dex (σ = 0.06 dex), completely compatible with MW halo stars. Stars from the ω Cen group, on the other hand, have an average [α/Fe] = 0.23 (σ = 0.06 dex), still consistent with the MW halo given the dispersion in both distributions, but less so than observed for the Kapteyn group.

Based on the run of α-element abundances, the stars of the Kapteyn group do not resemble the present-day dSph galaxies close to the MW, including the case of the Sagittarius dwarf, thus rejecting ω Centauri's parent galaxy as the origin for this moving group. The same can be said for the stars in the ω Cen group, but the somewhat lower overall [α/Fe] among these stars might be indicative of something different and does not allow us to discard completely an origin in a dSph galaxy.

Even though we do not find our samples of the Kapteyn and ω Cen groups to contain any second-generation stars, there are a few stars that appear special on their own, and we examine them separately here.

HD 151559, from our field sample, appears as a Na-rich and relatively O-poor star, compared with the other stars in our sample. This is one of the most metal-rich stars in the sample, with [Fe/H] ∼ −0.7, consistent with the metal-rich population found in ω Centauri, which is mainly composed of second-generation stars. In fact, HD 151559 falls in the same location of Figure 3 and of the right panel of Figure 4 where Ramírez et al. (2012) identified the first field dwarfs with O and Na abundances consistent with second-generation stars from GCs. One star out of 45 with both O and Na measurements amounts to just in between the 2%–3% of second-generation stars in the Galactic field recently found by Martell et al. (2011) and Ramírez et al. (2012). The He i EW measured for this star is below the maximum usually found in the literature for field stars. However, it is slightly larger than what is found for dwarfs and subdwarfs with [Fe/H] ≳ −1.0 dex (Takeda & Takada-Hidahi 2011). The magnesium and aluminum abundances for HD 151559 (Figure 5) are typical of MW halo stars. However, as discussed earlier, this alone does not say anything about its origin as a first- or second-generation star, since the Mg–Al anticorrelation is evident only in a few massive GCs, like ω Centauri, while other clusters with known different stellar generations are practically indistinguishable from the MW halo in its Mg and Al abundance content (see Figure 6 of Carretta et al. 2009b). It is the location of HD 151559 in the Na–O diagram that makes this star a firm candidate for a second-generation GC star in the Galactic halo.

CD-62 1346 and HD 222925 have an overabundance of Ba and resemble the Ba-rich stars typical of ω Centauri. However, even though their large He i EWs would be expected for helium-enhanced stars, their Na and O abundances are similar to field stars, which is not consistent with being second-generation stars from ω Centauri (Gratton et al. 2011; Marino et al. 2011). Therefore, the large He i EW in these cases might be due to chromospheric activity on those stars.

On the other hand, the high [Ba/Fe] values for these stars could be the result of mass transfer from an unseen AGB companion which changes the surface abundance of these stars. The most straightforward test for this possibility is to search for RV variations in time. For CD-62 1346 a RV of 125.74 ± 0.35 km s⁻¹ is found based on our optical spectrum, which is in excellent agreement with the value listed by Beers et al. (2000) of 127.0 ± 10 km s⁻¹. Despite CD-62 1346 not showing RV variations, its α and s-process element abundances are consistent with a CH-rich Ba-strong star in a binary system (see Pereira et al. 2012, and references therein). Therefore, the barium overabundance of this star is not representative of the Kapteyn group. HD 222925 is listed with V_hel = −45 ± 10 km s⁻¹ by Beers et al. while we found −38.64 ± 0.36 km s⁻¹. Due to the large uncertainties of the Beers et al. (2000) measurement, it is hard to conclude anything for this star. The abundances of yttrium and technetium are desirable to reject mass transfer from an AGB companion (Busso et al. 2001; Uttenthaler et al. 2011; Casey et al. 2014b, and references therein), but this kind of analysis is beyond the scope of this study.

Finally, HD 184266 has the largest value (>400 mÅ) for the He i EW of the whole sample. This star from our field sample is indistinguishable from the field halo stars with the same metallicity in its O, Na and Ba abundances. From visual inspection of its optical spectrum obtained with MIKE, the core of the chromospheric Ca ii lines is not found in emission, undermining the possibility of chromospheric activity as being responsible for the high He i EW (see Pasquini et al. 2011). The same is found for HD 222925, the star with the second largest He i EW (∼230 mÅ), also from our field star sample. A more detailed chromospheric/activity study of these stars is needed in order to determine whether these EWs are consistent with enhanced helium abundances.¹³