Atypical Mg-poor Milky Way Field Stars with Globular Cluster Second-generation-like Chemical Patterns

A number of recent observational studies have revealed that a handful of Milky Way (MW) field³⁵ stars may exhibit inhomogeneities in their light-element abundances (e.g., Carretta et al. 2010; Ramírez et al. 2012; Fernández-Trincado et al. 2016a; Martell et al. 2016; Recio-Blanco et al. 2017; Schiavon et al. 2017b) and neutron-capture element enhancements (e.g., Majewski et al. 2012; Hasselquist et al. 2016; Pereira et al. 2017), similar to those observed in the second-generation (SG)³⁶ population of globular clusters (GCs; e.g., Carretta et al. 2009a, 2009b; Mészáros et al. 2015; Carretta 2016; Pancino et al. 2017; Schiavon et al. 2017a; Tang et al.2017).

In this framework, the presence of stars with chemical anomalies in the Galactic field could be explained as the relics of tidally disrupted GCs (e.g., Majewski et al. 2012; Fernández-Trincado et al. 2016a and references therein), indicating that dissolved GCs could have deposited these eventually unbound stars into the main components of the MW (the bulge, the disk, and the halo; e.g., Carretta et al. 2010; Fernández Trincado et al. 2013; Kunder et al. 2014; Fernández-Trincado et al. 2015a, 2015b, 2016a, 2016b; Lind et al. 2015; Martell et al. 2016).

Despite the enormous progress that has recently been made in exploring abundance anomalies (e.g., C, N, Al) throughout the canonical components of the MW (e.g., Martell et al. 2016; Schiavon et al. 2017b), the distribution and properties of stars originally formed in GCs that are now part of the MW field are still not well understood. Therefore, the study of field stars with "polluted chemistry" opens a unique window to shed light on models that address the "mass budget" problem, stellar evolution models, and the phenomenon of multiple populations (MPs) in GCs (see Bastian & Lardo 2015; Ventura et al. 2016; Schiavon et al. 2017b). Here, we report the discovery of atypical MW field stars with SG-like chemical patterns from the Apache Point Observatory Galactic Evolution Experiment (APOGEE) survey.

Our sample was selected from the APOGEE survey, making use of Sloan Digital Sky Survey-IV (SDSS-IV) Data Release 13 (DR13; SDSS Collaboration et al. 2016; Majewski et al. 2017). APOGEE DR13 provides chemical and kinematical information of about 150,000 Galactic stars through the analysis of high-resolution (R ∼ 22,500) H-band λ = 1.51–1.69 μm spectra (Zasowski et al. 2013).

We focus our search on the low-metallicity regime (−1.8 < [Fe/H] < −0.7), where stars from the halo and thick disk are expected to dominate the Galactic metallicity distribution (Hawkins et al. 2015; Martell et al. 2016; C. R. Hayes et al. 2017, in preparation). We impose a minimal signal-to-noise ratio per pixel of 70 to ensure good-quality spectra. In order to identify abundance anomalies in MW field stars, we proceed as follows:

From our initial sample (4611 stars) we selected stars with SG-like chemical patterns in the [Mg/Fe] versus [Al/Fe] plane by means of a clustering analysis. This is done using a k-means clustering approach as described in Ivezić et al. (2014), with three different centroids in two-dimensional chemical space ([Mg/Fe], [Al/Fe]): (i) the SG stars from Galactic GCs ($+0.1,+0.7$); (ii) the FG stars in Galactic GCs ($+0.15,-0.2$); and (iii) the Galactic thick disk stars ($+0.25,+0.2$). Furthermore, we extended the limits on the Al distribution provided by the k-means analysis for SG-like stars using generous Al cuts ([Al/Fe] ≳ +0.1) and searched for SG-like stars, omitting the carbon-rich stars [C/Fe] ≳ +0.15 (Schiavon et al. 2017b), which exhibit anomalous chemical abundance patterns as observed in SG GC stellar populations. All the raw data used in this Letter are available in a public repository.³⁷

Figure 1(a) shows the locus occupied by our SG-like candidates, which is located above the dashed gray line that was derived according to the k-means algorithm. Stars from Galactic GCs of similar metallicity (Mészáros et al. 2015) and the N-rich field stars of Martell et al. (2016) and Schiavon et al. (2017b) are also indicated in the figure for illustration. Indeed, 10 of the N-rich stars reported by Schiavon et al. (2017b) are situated in the locus of SG-like stars found by the k-means algorithm.

Zoom In Zoom Out Reset image size

After applying the criteria cited above, we have a subsample of 260 stars, from which 58.5% (152/260) are known stars from clusters and other anomalous stars previously reported in the literature (Mészáros et al. 2015; Fernández-Trincado et al. 2016b; Schiavon et al. 2017b; Tang et al. 2017) and 28.5% (74/260) have no significant N overabundances (see Section 3) and were rejected.

To discard false positives in the remaining 34 stars, the most relevant atomic (Al, Mg, Si, and Ni) and molecular (CO, CN, and OH) spectral features in the H-band were visually inspected to ensure that the final APOGEE spectra are of good quality (e.g., not critically affected by detector persistence, proper continuum normalization, telluric- and sky-line correction, etc.), to provide reliable chemical abundances. We end with a final sample comprising 11 stars (Table 1).

We have analyzed up to nine chemical elements that are typical indicators of the presence of SG stars in GCs (C, N, O, Mg, Al, Si, Ni, Na, and Fe). The APOGEE DR13 does not provide reliable N abundances for most of our potential candidates because they show very strong CN lines, falling near the high-N edge of the grid and consequently flagged as "GRIDEDGE_BAD" in DR13 (except 2M02491285+5534213 with [N/Fe] = +0.67; see Figure 1(c)).

In order to provide a consistent chemical analysis, we re-determine the chemical abundances by means of a line-by-line analysis. The chemical abundances have been derived assuming as input the effective temperature (T_eff) and metallicity as derived by the APOGEE Stellar Parameter and Chemical Abundances Pipeline (ASPCAP; García Pérez et al. 2016). However, we do not adopt the surface gravity (log g) provided by ASPCAP, as it is affected by a systematic effect that overestimates the log g values (J. A. Holtzman et al. 2017, in preparation). We estimate surface gravity from 10 Gyr PARSEC (Bressan et al. 2012) isochrones (10 Gyr is the typical age of Galactic GCs; Harris 2010). The linelist used in this work is the latest internal DR13 atomic/molecular linelist (linelist.20150714), and the line-by-line analysis was done using the 1D spectral synthesis code Turbospectrum (Alvarez & Plez 1998) and MARCS model atmospheres (Gustafsson et al. 2008). In particular, a mix of heavily CN-cycle and α-poor MARCS models were used. The same molecular lines adopted by Smith et al. (2013) and Souto et al. (2016) were employed to determine the C, N, and O abundances. Examples for a portion of the observed APOGEE spectra (spectral region covering CN, Mg, and Al lines) are shown in Figure 2 for our 11 anomalous stars. Table 1 lists the final set of atmospheric parameters and chemical abundances for each star obtained through ASPCAP DR13 (first line) and the line-by-line synthesis calculations adopting log g from theoretical isochrones and using the tools mentioned above (second line).

Zoom In Zoom Out Reset image size

We find the differences in the star-to-star abundances between ASPCAP DR13 and our manual analysis to be small, Δ[Mg/Fe] ≲ +0.2, Δ[Al/Fe] ≲ +0.15, Δ[O/Fe] ≲ +0.2, Δ[Si/Fe] ≲ +0.15, and Δ[Ni/Fe] ≲ +0.15, generally overlapping with our internal errors. It is important to note that these discrepancies do not affect the main conclusion of this work, i.e., both line-to-line abundances and DR13 abundances indicate that these stars are N-rich and Al-rich. Mg abundances are usually lower in the manual analysis compared with ASPCAP, a result already found in similar type of SG-like field stars (Fernández-Trincado et al. 2016a). We note that Na abundances are more discrepant between DR13 and our manual analysis. As the Na lines are usually weak (especially in the most-metal-poor stars; [Fe/H] < −1.0), the uncertainty in the Na abundance is strongly modulated by the uncertainty in the continuum location. ASPCAP uses a global fit to the continuum in three detector chips independently, while we place the pseudo-continuum in a region around the lines of interest. We believe that our manual method is more reliable, as it avoids possible shifts in the continuum location due to imperfections in the spectral subtraction along the full spectral range. This way, our manual analysis shows the Na-rich nature of the SG-like candidates.

We use the galactic dynamic software GravPot16³⁸ (Model 4 in Fernández-Trincado et al. 2016a) to predict the trajectories for five stars (Table 2), from which the space velocity and position vectors can be fully resolved.

Table 2.  Variations between 2MASS and DENIS Magnitudes and Radial Velocities ($\sigma {RV}$) over the Period of the APOGEE Observations

APOGEE ID	${K}_{2\mathrm{MASS}}-{K}_{\mathrm{DENIS}}$	Nvisits	$\sigma {RV}$	Median rperi	Median rapo	Median Zmax	Median e
	(mag)		(km s−1)	(kpc)	(kpc)	(kpc)
2M17535944+4708092	⋯	3	0.21	${3.84}_{-2.4}^{+4.3}$	${21.54}_{-5.5}^{+48}$	${18.79}_{-7.6}^{+43.4}$	${0.76}_{-0.19}^{+0.16}$
2M17585001-2338546	−0.064	1	⋯	⋯	⋯	⋯	⋯
2M17350460-2856477	0.134	2	0.23	⋯	⋯	⋯	⋯
2M12155306+1431114	⋯	13	0.13	${4.078}_{-2.9}^{+3.8}$	${17.6}_{-1.9}^{+29}$	${16.04}_{-1.9}^{+17.41}$	${0.69}_{-0.15}^{+0.2}$
2M16062302-1126161	−0.071	4	0.24	${1.15}_{-0.76}^{+0.49}$	${5.7}_{-0.43}^{+0.33}$	${3.32}_{-0.43}^{+0.43}$	${0.66}_{-0.09}^{+0.19}$
2M17454705-2639109	−0.056	1	⋯	⋯	⋯	⋯	⋯
2M17492967-2328298	0.117	2	0.10	⋯	⋯	⋯	⋯
2M17534571-2949362	⋯	2	0.07	${0.92}_{-0.66}^{+0.89}$	${6.18}_{-0.97}^{+2.08}$	${0.74}_{-0.43}^{+1.74}$	${0.73}_{-0.12}^{+0.18}$
2M11462612-1419069	0.048	4	0.08	⋯	⋯	⋯	⋯
2M17180311-2750124	0.039	2	0.08	${0.167}_{-0.12}^{+0.56}$	${5.40}_{-1.59}^{+1.0}$	${2.27}_{-0.68}^{+0.79}$	${0.94}_{-0.19}^{+0.04}$
2M02491285+5534213	⋯	3	0.20	⋯	⋯	⋯	⋯

Note. Columns 5, 6, 7, and 8 show the median perigalactic distance, the median apogalactic distance, the median maximum distance from the galactic plane, and the median eccentricity, respectively. The orbital eccentricity is defined as $e=({r}_{\mathrm{apo}}-{r}_{\mathrm{peri}})/({r}_{\mathrm{apo}}+{r}_{\mathrm{peri}})$, with r_apo and r_peri as the perigalactic and apogalactic radii of the orbit, respectively. The orbital elements given here are estimates from Monte Carlo simulations of 10⁵ orbits.

Download table as:  ASCIITypeset image

To construct the stellar orbits, we employed radial velocities derived from APOGEE DR13, proper motions from UCAC-5 (Zacharias et al. 2017), and APOGEE distance estimates from Santiago et al. (2016) and F. Anders (2017, in preparation). The orbital elements are listed in Table 2.

All five stars indeed lie on very eccentric orbits (e > 0.65) passing through the Galactic bulge, reflecting a potentially unusual origin in the MW.

In particular, two stars (2M17535944+4708092 and 2M12155306+1431114) have relatively high metallicity ([Fe/H] ∼ −0.8) and may reach distances of up to ${Z}_{\max }\,\sim 17$ kpc above the Galactic plane.

These orbital properties (together with the unusually low levels of Mg observed in the most-metal-rich stars) may support our speculated scenario discussed below, in which these atypical stars may have an extragalactic origin.

The main finding of this work is the discovery of 11 atypical MW field red giant stars with SG GC-like abundance patterns, i.e., with strong enrichments in N, Na, Si, and Al, accompanied by decreased abundances of C, O, Ni, and Mg. Figure 1(b) shows that most of the new chemically anomalous stars exhibit significantly lower [Mg/Fe] ratios (at [Fe/H] ≳ −1.0) as compared to Galactic disk stars (at the same metallicity) and the N-rich halo and bulge stars (e.g., Martell et al. 2016; Schiavon et al. 2017b). This suggests that the vast majority of our stars have an unusual origin. The exceptions are the two most-metal-poor stars ([Fe/H] ≲ −1.4), which display higher [Mg/Fe] ratios similar to the "canonical halo." Their [Al/Fe] and [N/Fe] ratios, however, are significantly higher than those of the bulk of MW field stars (Figures 1(a) and (c)), indicating that they may be SG stars originally formed from material that was chemically enriched in GCs (Martell et al. 2016; Schiavon et al. 2017b). For example, the measured abundances are in good agreement with the pollution expected by massive AGB stars at metallicity lower than [Fe/H] < −1.4 (Ventura et al. 2016; F. Dell'Agli et al. 2017, in preparation).

Interestingly, the most-metal-rich ([Fe/H] ≳ −1.2) and atypical Mg-poor stars appear to belong to two groups, according to their Fe abundance (see Figure 1). A first group, only two stars with −1.2 ≲ [Fe/H] ≲ −1.0, exhibit Mg depletion more or less consistent (within the errors) with the Mg abundances typically observed in Galactic GCs of similar metallicities (Mészáros et al. 2015). The second group (seven stars), however, displays similar Mg depletion (Figure 1), but at higher metallicities ([Fe/H] > −1.0). This Mg-deficiency ([Mg/Fe] ≲ 0)—coupled with strong N and Al enrichment ([N,Al/Fe] ≳ +0.5)—is at odds with present observations of SG stars in Galactic GCs of similar metallicities (Figure 1).³⁹ In addition, Figure 1 shows that this Mg-deficiency is not seen in the vast majority of N-rich bulge stars of similar metallicity (Schiavon et al. 2017b); only one N-rich bulge star displays a chemical pattern identical to the atypical stars reported here (Figure 1). A total of six atypical sample stars are seen to lie toward the bulge, but it is not clear if they could (or not) be some kind related to the latter N-rich bulge population.

Could these atypical stars be chemically tagged as migrants from dwarf galaxies? We find this possibility unlikely because our stars display [Al/Fe] much higher than observed in dwarf galaxy stellar populations today (e.g., Shetrone et al. 2003; Hasselquist et al. 2017). However, these stars could be former members of a dwarf galaxy (with intrinsically lower Mg) polluted by a massive AGB star in a binary system, which could produce the chemical pattern observed. Such an exotic binary system seems to be unlikely. Indeed, no star in our sample exhibits significant photometric and/or radial velocity variability (see Table 2). Follow-up observations (e.g., more radial velocity data) would confirm/disprove the binary hypothesis.

Recently, Ventura et al. (2016) have reported a remarkable agreement between the APOGEE Mg–Al anticorrelations (two elements sensitive to the metallicity of the GC polluters) observed in Galactic GCs (−2.2 ≲ [Fe/H] ≲ −1.0) and the theoretical yields from massive AGB stars (m-AGBs). This further supports the idea that SG-GC stars formed from the winds of m-AGBs, possibly diluted with pristine gas with the same chemical composition of the FG stars (see also Renzini et al. 2015). At higher metallicities −1 < [Fe/H] < −0.7, however, the maximum Al spread (with respect to the FG) expected from the ejecta of m-AGBs is in the range +0.2 < Δ(Al) < +0.5 (Ventura et al. 2016; F. Dell'Agli et al. 2017, in preparation), but only a modest Mg depletion is expected. The high Al observed ([Al/Fe] ≳ +0.6) in the atypical stars at these metallicities could be explained under the m-AGBs pollution framework if they are earlier SG members of dissolved GCs (see Schiavon et al. 2017b) where the FG stars formed with higher levels of Al. The FG stars in metal-rich ([Fe/H] ≳ −1.0) Galactic GCs such as M107, M71, 47 Tuc, and NGC 5927 (Mészáros et al. 2015; Pancino et al. 2017) are known to be formed with a higher Al (compared to a purely solar-scaled mixture), but both FG and SG stars exhibit similarly high Mg abundances—with no significant spread between the two stellar generations, as predicted by the m-AGBs self-enrichment scenario (Ventura et al. 2016; F. Dell'Agli et al. 2017, in preparation).

Therefore, the chemical composition of our atypical metal-rich stars, particularly the observed Al overabundances coupled with low Mg, cannot be explained by invoking pollution from m-AGBs alone (formed with a solar-scaled or an α-enhanced mixture). A possible explanation for these chemical anomalies is that these stars escaped from GCs whose FG stars formed with a chemical composition enriched in Al but with a lower Mg content in comparison with the standard solar-scaled or α-enhanced mixture. This could be obtained if we hypothesize that the gas cloud from which the GC formed was mainly polluted by SN explosions of stars of about ∼20−30 M_⊙, characterized by medium or large rotation rates during their life, according to the most recent yields by M. Limongi & A. Chieffi (2017, in preparation). Under these conditions the gas ejected is expected to be slightly enriched in Al, but is Mg-poor. If the FG stars formed with this chemistry, then subsequent pollution from m-AGBs would form SG stars with the same chemical composition of the atypical Mg-poor SG-like stars reported here.

[Mg/Fe] (or [Mg/α]) from high-resolution integrated-light spectroscopic observations in extragalactic GCs—even with average metallicities similar to our atypical Mg-poor stars—is generally lower than in Galactic GCs with a similar metallicity (e.g., Pancino et al. 2017). A low [Mg/Fe] ratio coupled with high Al (when available) is also observed in some extragalactic GCs (e.g., in M31 and LMC GCs; see, e.g., Colucci et al. 2009, 2012). At present, possible explanations for the low Mg content in some extragalactic GCs include both internal and external effects, which could also work simultaneously (e.g., Pancino et al. 2017). The internal effect is linked to the particular formation and chemical evolution of a given GC (e.g., NGC 2419), while the external effect is related to the specific chemical evolution of their host galaxies.

In short, the unique Mg-deficiency of the discovered atypical metal-rich stars with SG-like chemical patterns (as well as their orbital properties) suggest that these stars may have an extragalactic origin, e.g., they could be former members of dissolved extragalactic GCs, the remnants of stellar systems accreted a long time ago by our Galaxy. This finding should encourage future dedicated searches (e.g., with ongoing massive spectroscopic surveys like APOGEE-2, Gaia-ESO, etc.) of chemically atypical Galactic stars, something that would represent a major advancement to understanding the formation and evolution of our own Galaxy.

J.G.F.-T. would like to dedicate this work to the memory of his father, José Gregorio Fernández Rangel (1959–2017).

We thank Paolo Ventura and Marco Limongi for useful discussions. We are grateful to the referee for a prompt and constructive report. J.G.F.-T. is especially grateful for the technical expertise and assistance provided by the Instituto de Astrofísica de Canarias (IAC). J.G.F.-T., D.G., and B.T. gratefully acknowledge support from the Chilean BASAL Centro de Excelencia en Astrofísica y Tecnologías Afines (CATA) grant PFB-06/2007. S.V. gratefully acknowledges the support provided by Fondecyt reg. No. 1170518. D.A.G.H. was funded by the Ramón y Cajal fellowship number RYC-2013-14182. D.A.G.H., O.Z., and F.D. acknowledge support provided by the Spanish Ministry of Economy and Competitiveness (MINECO) under grant AYA-2014-58082-P. S.M. has been supported by the Premium Postdoctoral Research Program of the Hungarian Academy of Sciences, and by the Hungarian NKFI grants K-119517 of the Hungarian National Research, Development and Innovation Office. D.M. is also supported by FONDECYT No. 1170121 and by the Ministry of Economy, Development, and Tourism's Millennium Science Initiative through grant IC120009, awarded to the Millennium Institute of Astrophysics (MAS). E.M, B.P., and A.P.V. acknowledge support from UNAM/PAPIIT grants IN105916 and IN114114. A.R.L. thanks partial financial support from the FONDECYT REGULAR project 1170476.

Funding for the GravPot16 software has been provided by the Centre national d'études spatiales (CNES) through grant 0101973 and the UTINAM Institute of the Université de Franche-Comté supported by the Region de Franche-Comté and Institut des Sciences de l'Univers (INSU).

Funding for the Sloan Digital Sky Survey IV has been provided by the Alfred P. Sloan Foundation, the U.S. Department of Energy Office of Science, and the Participating Institutions. SDSS-IV acknowledges support and resources from the Center for High-Performance Computing at the University of Utah. The SDSS web site is www.sdss.org. SDSS-IV is managed by the Astrophysical Research Consortium for the Participating Institutions of the SDSS Collaboration including the Brazilian Participation Group, the Carnegie Institution for Science, Carnegie Mellon University, the Chilean Participation Group, the French Participation Group, Harvard-Smithsonian Center for Astrophysics, Instituto de Astrofìsicade Canarias, The Johns Hopkins University, Kavli Institute for the Physics and Mathematics of the Universe (IPMU)/University of Tokyo, Lawrence Berkeley National Laboratory, Leibniz Institut für Astrophysik Potsdam (AIP), Max-Planck-Institut für Astronomie (MPIA Heidelberg), Max-Planck-Institut für Astrophysik (MPA Garching), Max-Planck-Institut für Extraterrestrische Physik (MPE), National Astronomical Observatory of China, New Mexico State University, New York University, University of Notre Dame, Observatório Nacional/MCTI, The Ohio State University, Pennsylvania State University, Shanghai Astronomical Observatory, United Kingdom Participation Group, Universidad Nacional Autónoma de México, University of Arizona, University of Colorado Boulder, University of Oxford, University of Portsmouth, University of Utah, University of Virginia, University of Washington, University of Wisconsin, Vanderbilt University, and Yale University.