Chemical Tagging N-rich Field Stars with High-resolution Spectroscopy

We started the selection of observable targets from our LAMOST N-rich field star sample (Papers I and II), which are mostly located in the northern sky. Fifteen stars with DEC <10° and a reasonable airmass were selected. Our follow-up observations were taken using the MIKE (Magellan Inamori Kyocera Echelle, Bernstein et al. 2003) spectrograph on the Magellan Clay telescope (Shectman & Johns 2003) at Las Campanas Observatory, Chile. The data were taken in two runs, 2019/07 and 2020/02. The blue and red detectors were used simultaneously to cover 3200 < λ < 5000 Å (blue side) and 4900 < λ < 10000 Å (red side). We used the 0.7'' slit and 2 × 2 spatial on-chip binning to achieve nominal spectral resolution of 35k/28k on the blue side and red side, respectively. Depending on the observation schedule and the brightness of the star, we took several exposures to ensure a signal-to-noise ratio (S/N) > 50 over most wavelengths of the red detector (Table 1).

Table 1. Stellar Parameters

Id	R.A.	Decl.	Teff	$\mathrm{log}g$	[Fe/H]	Gmag	Microturbulence	RV	Airmass	Exposure Time
			(K)	(dex)	(dex)	(mag)	(km s−1)	(km s−1)		(s)
9 (II)	58.221019	7.203154	4312	0.65	−1.34	13.63	1.78	69.0	1.42	4800
38 (II)	258.215699	8.130489	4604	0.73	−1.62	12.57	2.04	−146.0	1.90	3600
47 (II)	214.054033	−2.145359	4953	2.14	−1.38	12.06	1.95	66.5	1.80	3600
58 (II)	197.656039	−6.979531	5213	2.84	−1.45	14.45	1.78	−27.0	1.10	5400
62 (II)	122.120480	1.946907	5020	2.36	−1.42	12.87	1.56	147.0	1.17	5400
65 (II)	233.551930	0.326274	4806	1.69	−1.63	12.76	1.45	−72.5	1.57	3600
67 (I)	317.852325	−2.385546	5265	2.12	−0.97	13.06	2.46	−5.0	1.14	3600
69 (I)	150.981537	1.298311	5129	3.02	−1.07	12.22	1.26	50.0	1.12	3600
80 (I)	134.988876	1.055260	4800	1.37	−1.88	13.79	1.35	179.0	1.00	5400
82 (I)	195.094070	−7.636771	5017	1.90	−1.06	12.59	1.60	−6.0	1.50	5400
86 (I)	239.791901	10.020472	5149	2.91	−0.98	14.10	1.37	−141.0	1.47	3600
88 (I)	244.844238	9.523706	4879	2.31	−1.02	13.16	1.51	−127.0	1.88	3600
94 (I)	166.054535	8.444023	5161	2.88	−1.18	14.44	1.33	51.0	2.00	3600
97 (I)	222.293640	10.045577	4762	2.20	−1.25	13.12	1.34	−31.0	1.49	3600
98 (I)	199.145416	10.957638	4529	0.77	−1.07	10.59	2.03	84.0	1.30	3600

Note. The Id is the corresponding id in Paper II (Table 3), with (I) and (II) indicate that they are from Paper I and Paper II, respectively.

Download table as:  ASCIITypeset image

We reduced the observational data using the CarPy package (Kelson et al. 2000; Kelson 2003). The observed spectral images were bias-subtracted, flat-field corrected, wavelength calibrated, scatter-light, and sky subtracted. The multiorder reduced spectra were then merged into a single spectrum.

Stellar parameters and chemical abundances were analyzed using the Brussels Automatic Stellar Parameter (BACCHUS) code (Masseron et al. 2016). We briefly describe the basic procedures here; readers are referred to Masseron et al. (2016) for detailed descriptions.

The equivalent widths (EW) of Fe i and Fe ii absorption lines are first measured. To derive stellar parameters, several stellar parameters are attempted before they converge: T_eff is determined by obtaining a null trend for the abundance of iron lines against excitation potential (see Figure 11 in the Appendix); $\mathrm{log}g$ is obtained by the ionization equilibrium between Fe i and Fe ii lines; metallicity is obtained by the mean of individual line abundances relative to the solar value; and microturbulence velocity is determined by obtaining a null trend for the abundance of iron lines against equivalent widths. As an initial guess, we used photometric T_eff calculated from 2MASS photometry (González Hernández & Bonifacio 2009) and $\mathrm{log}g$, [Fe/H] from LAMOST suggested values. Our derived stellar parameters using BACCHUS are shown in Table 1. Note that most of our stars show $\mathrm{log}g$ greater than 2. Because the red clump feature is located at $\mathrm{log}g\sim 2$ Paper II, our sample stars are mostly RGB stars.

To derive chemical abundance for a given absorption line, a sigma-clipping is first applied on the selected continuum points around the targeted line, then a linear fit is used for the remaining points as the continuum. Therefore, the code can detect significantly bad fits, like a sudden drop in the spectrum due to bad pixels in the detectors. Observed spectra and model spectra are compared in four different methods to determine abundances: χ² minimization, line intensity, equivalent width, and spectral synthesis (see Figure 12 in the Appendix). Furthermore, the code gives each of them a flag to indicate the estimation quality. When multiple transition lines are detected for a given element, the lines that satisfy the following are chosen: (1) λ > 5000 Å (for better S/N, see Figure 10 in the Appendix); (2) all four determination methods are flagged as good (=1); and (3) S/N > 40. Then, the mean value of the chemical abundances of different absorption lines from χ² minimization is given as the estimated abundance for the given element. If no line meets the above criteria, the abundance is not estimated. The derived chemical abundances of our targets are given in Table 2.

In this work, we analyzed the internal errors by propagating the typical errors in T_eff, $\mathrm{log}g$, and [Fe/H]. The typical errors are set as ΔT_eff = 50 K, ${\rm{\Delta }}\mathrm{log}g=0.1\,\mathrm{dex}$, and Δ[Fe/H] = 0.05 dex. The total estimated error is thus calculated as ${\sigma }_{\mathrm{tot}}\,={\left({\left({\sigma }_{{\rm{\Delta }}{T}_{\mathrm{eff}}}\right)}^{2}+{\left({\sigma }_{{\rm{\Delta }}\mathrm{log}g}\right)}^{2}+{\left({\sigma }_{{\rm{\Delta }}[\mathrm{Fe}/{\rm{H}}]}\right)}^{2}\right)}^{1/2}$. The abundance errors of a typical star (star #88 from Paper II) are shown in Table 3 as an example.

Table 3. Errors of Chemical Abundances Propagated from Atmospheric Parameters for Star #88

Element	Abundance	ΔTeff = 50(K)	${\rm{\Delta }}\mathrm{log}(g)=0.1$ (dex)	Δ[Fe/H] = 0.05 (dex)	σtot
Δ([O/Fe])	0.56	0.00	0.03	0.03	0.04
Δ([Na/Fe])	0.23	0.03	0.00	0.08	0.09
Δ([Mg/Fe])	0.27	0.03	0.00	0.00	0.03
Δ([Al/Fe])	0.14	0.02	0.00	0.02	0.03
Δ([Si/Fe])	0.26	0.01	0.01	0.02	0.03
Δ([S/Fe])	⋯	⋯	⋯	⋯	⋯
Δ([Ca/Fe])	0.23	0.04	0.00	0.00	0.04
Δ([Sc/Fe])	0.20	0.00	0.03	0.02	0.04
Δ([Ti/Fe])	0.22	0.04	0.01	0.01	0.04
Δ([V/Fe])	−0.05	0.07	0.00	0.00	0.07
Δ([Cr/Fe])	−0.16	0.03	0.02	0.04	0.06
Δ([Mn/Fe])	−0.51	0.08	0.01	0.01	0.08
Δ([Co/Fe])	−0.17	0.06	0.01	0.01	0.07
Δ([Ni/Fe])	−0.09	0.03	0.01	0.03	0.04
Δ([Y/Fe] )	0.08	0.04	0.04	0.05	0.07
Δ([Zr/Fe])	⋯	⋯	⋯	⋯	⋯
Δ([Ba/Fe])	0.46	0.01	0.05	0.06	0.08
Δ([La/Fe])	0.63	0.01	0.04	0.01	0.05
Δ([Ce/Fe])	0.32	0.01	0.02	0.03	0.04
Δ([Nd/Fe])	0.49	0.03	0.04	0.07	0.09
Δ([Eu/Fe])	⋯	⋯	⋯	⋯	⋯

Download table as:  ASCIITypeset image

The anticorrelation between Na and O is observed in almost all GCs (Carretta et al. 2009b, 2010b; Gratton et al. 2012), which is arguably the most characteristic MP property of GC member stars. The Na–O anticorrelation is suggested to originate from the CNO cycle and the NeNa cycle activated during lower temperature H-burning. If our N-rich field stars show compatible Na and O abundances as the GC enriched populations, it would be a supportive evidence for their GC origin scenario. For the reference of chemical abundances of GC stars, we used the observed data of 19 GCs from Carretta et al. (2009b). Based on the Na and O abundances, they separated members of each GC into three components, the P (primordial) component, the I (intermediate) component, and the E (extreme) component. FG stars (or P component) are assumed to have similar O and Na abundances as field stars of the same metallicity, while I and E components are all considered to be SG stars.

Following the convention to measure O abundances (e.g., Carretta et al. 2009b), we only used [O I] λ λ 6300, 6363Å forbidden lines to minimize the effect of nonlocal thermodynamic equilibrium (NLTE) correction. If no measurable line is found, the mean value from χ² minimization is given as the upper limit. The Na and O abundances are shown in Figure 1, where GC stars from Carretta et al. (2009b) are shown as background dots and contours. In order to match the metallicities between our N-rich field stars and GCs, only GCs with − 1.8 ≤ [Fe/H] ≤ − 1.0 are used. Though a clear anticorrelation between Na and O with large spread is found for GC stars, the separation between FG and SG stars is not clear-cut. Seven stars have measurable O abundances. Most N-rich field stars basically show consistent Na and O abundances with those in GCs, except for two stars: one with [O/Fe ] ∼ 1 and the other with [Na/Fe ] ∼ 1. Based on their measurable Na abundances, most N-rich field stars are located in the transition region between FG and SG stars, thus it is difficult to verify their SG origin.

Zoom In Zoom Out Reset image size

Mg–Al anticorrelation is another common feature found in low metallicity GCs ([Fe/H] < − 1.0; Pancino et al. 2017), which is suggested to be the consequence of the Mg–Al cycle activated at higher core temperature during H-burning (Arnould et al. 1999). Our N-rich field stars are shown as red symbols in Figure 2. We note that the red pentagons are chemical abundances derived in Paper II using APOGEE spectra. In the background, GC stars from Carretta et al. (2009a) are separated into FG (orange) and SG (blue) stars as previously mentioned.

Zoom In Zoom Out Reset image size

Though GC FG and SG stars are widespread and even overlap in [Al/Fe] distribution, FG stars seldomly exceed [Al/Fe] ∼ 0.5 (excluding upper limits). On the other hand, GC SG stars may show low [Al/Fe] down to 0, and thus a lower [Al/Fe] cannot exclude the possibility that the star belongs to SG. In this sense, almost half of the stars our N-rich field star sample show [Al/Fe] > 0.5, and thus are probable SG stars. Those stars with [Al/Fe] > 0.5 also have compatible [Mg/Fe] to ensure that they are covered by the Mg–Al anticorrelation of GC Al-enhanced stars. There are two extremely Mg-depleted N-rich field star ([Mg/Fe] < − 0.4), but one of them has been reported in Paper II and Fernández-Trincado et al. (2019). The Mg-depleted N-rich field stars were also discussed in Fernández-Trincado et al. (2016, 2017). The Mg depletion in N-rich field stars is rare, it would be interesting to know if these Mg-depleted N-rich field stars have gone through additional nucleosynthesis. Meanwhile, the other half of N-rich field stars ([Al/Fe] < 0.5) show Mg and Al abundances consistent with GC FG/SG stars. Though chemically being less distinguishable from FG stars based on their Mg and Al abundances, we cannot rule out the possibility that lower [Al/Fe] N-rich field stars are SG stars, as more metal-rich GCs tend to show similar Al abundances for FG and SG stars (Pancino et al. 2017). Furthermore, using APOGEE GC stars (Mészáros et al. 2020), we found that lower Al abundance ([Al/Fe] < 0.5) N-rich field stars tend to be more metal-rich in the sample (Figure 3), tentatively agreeing with the aforementioned GC behavior.

Zoom In Zoom Out Reset image size

The α-elements are mainly generated in massive stars through type II supernovae (SNe II), consequently their recycle timescale in the interstellar medium is much smaller than that of iron, which is mainly produced in type Ia supernovae (SNe Ia). As a result, the [α/Fe] abundance is initially enhanced, and starts to decline with [Fe/H] after the SN Ia explosion rate reaches a maximum (Matteucci & Greggio 1986). The decline point, or the knee in [α/Fe] versus [Fe/H] trend, is related to star formation rate and thus galaxy mass: the less massive the galaxy is, the more metal-poor the [α/Fe] turnover is. O and Mg are commonly classified as hydrostatic α-elements, while Si, Ca, and Ti are classified as explosive α-elements (Woosley & Weaver 1995), implying different nucleosynthetic processes.

The [Mg/Fe]–[Fe/H] relation is shown in the upper left panel of Figure 4. In the background, we also include MW halo and disk stars, dwarf galaxy (Sculptor and Sagittarius) stars, and APOGEE GC stars (violin shaped symbols). ⁸ Each violin shaped symbol corresponds to one GC. The center value and min–max values are determined by its GC members selected from the APOGEE survey (e.g., Mészáros et al. 2020; Fernández-Trincado et al. 2020). [Mg/Fe] of our N-rich field stars are inside the min–max range of APOGEE GCs of similar metallicity, which does not contradict with the hypothesis that N-rich field stars come from GCs. However, in situ GCs (GCs identified as main disk or main bulge) and accreted GCs are largely overlapped especially in our chosen metallicity range ( − 1.8 < [Fe/H] < − 1.0), which prevents us from distinguishing their in situ or accreted GC origin. Meanwhile, our N-rich field stars are mostly covered by both data points of MW stars and dwarf galaxies, indicating a mixture of in situ stars and extragalactic stars. This is consistent with their kinematics in Paper II.

Zoom In Zoom Out Reset image size

As the atomic number increases, Si, Ca, and Ti become more stable against H-burning. Si abundance is suggested to vary by the "Si-leakage" during the Mg–Al cycle when the core temperature reaches as high as T ∼ 65 × 10⁶ K (Carretta et al. 2009a). Most GCs do not show significant variation in Ca (e.g., Carretta et al. 2010a), except for several massive GCs with possible iron spread (Carretta & Bragaglia 2021). No significant Ti variation among GC members is reported to the best of our knowledge.

Interestingly, Horta et al. (2020) reported that in situ GCs may have higher [Si/Fe] compared to accreted GCs for [Fe/H] > − 1.5. The [Si/Fe] difference between two groups of GCs becomes less distinguishable as metallicity decreases. The background MW stars and dwarf galaxy stars (Sculptor and Sagittarius) of Figure 4 (bottom left panel) support the statement of Horta et al. (2020) for field stars; though the threshold metallicity can be only vaguely located at −1.5 < [Fe/H] < −1.0. Upper right panel and bottom right panel of Figure 4 shows the Ca and Ti abundance distributions of our N-rich field stars, along with background MW stars, dwarf galaxy stars, and GC star distribution (violin shaped symbols), respectively. Jönsson et al. 2020 warned against using APOGEE-derived Ti abundances, since the Ti I and Ti II abundances from the APOGEE pipeline may have unknown defects or large scatters, so [Ti/Fe] for APOGEE GCs are not shown in this work.

Looking at Mg, Si, Ca, and Ti abundances versus [Fe/H] simultaneously, interesting results emerge: (1) The α-element difference between in situ (MW) stars and accreted (dwarf galaxy) stars is most significant for Ti, compared to Mg, Si, and Ca. It is promising to use [Ti/Fe] for future works to distinguish accreted stars. (2) The [α/Fe] from the APOGEE survey (violin shaped symbols) cover most of the N-rich field stars, again supporting their GC origin. (3) Seven metal-rich ( − 1.25 ≤ [Fe/H] ≤ − 0.95) N-rich field stars show abundances more consistent with MW stars, indicating a higher possibility of in situ origin. (4) At the lower end of the [α/Fe] distribution, star #9 (labeled with large red circle in Figure 4) shows consistently low Mg, Si, Ca, and Ti abundances similar to stars in dwarf galaxies, although it shows similar [Al/Fe] as MW disk and halo stars (Figure 3). However, Al can be enhanced in GCs without altering much of its α-abundances. Is it possible that this star has an extragalactic GC origin? We will further discuss this with more chemical information below.

Though Type Ia SNe, runaway deflagration obliterations of white dwarfs, have a signature more tilted toward the iron-peak group (Nomoto et al. 1997), the solar composition of the iron-peak elements are in fact a heterogeneous combination of both Type Ia SNe and core collapse Type II SNe (Woosley & Weaver 1995). As dwarf galaxies and MW have different star formation timescale, this discrepancy may manifest itself in iron-peak elements. Figure 5 show that MW stars (black dots and magenta squares) and dwarf galaxy stars (blue circles) have appreciable different distributions in [Sc/Fe], [V/Fe], and [Co/Fe] versus [Fe/H] graphs over the given metallicity range, especially in the metal-rich part ([Fe/H] ≥ − 1.25). For these three elements, more metal-rich N-rich field stars ( − 1.25 ≤ [Fe/H] ≤ − 0.95) show distributions more similar to MW stars, confirming a higher possibility of in situ origin (Section 3.3). For [Cr/Fe], [Mn/Fe], and [Ni/Fe], MW stars and dwarf galaxy stars show less distinguishable differences. The discrepancies in iron-peak element of MW stars and dwarf galaxies are not evident in the metal-poor part (especially for − 1.25 < [Fe/H] < − 1.5) due to the lack of data. However, if a linear relation of iron-peak abundances and metallicity is assumed for both MW (halo and disk) and dwarf galaxies, the dwarf galaxies would have lower iron-peak abundances. Therefore, star #9 with low α-element abundances also shows lower abundances in [Sc/Fe], [V/Fe], and [Co/Fe], agreeing with dwarf galaxies, which supports its possible extragalactic GC origin.

Zoom In Zoom Out Reset image size

Elements heavier than iron are generated via the neutron-capture processes. Depending on the relative speed of neutron capture compared to β-decay, the neutron-capture processes are divided into rapid (r-) and slow (s-) ones. The main s-process elements are synthesized by AGB stars during thermal pulsations (Busso et al. 2001; Karakas & Lattanzio 2014). The astrophysical sites of producing r-process elements have been debated over the past sixty years (e.g., Thielemann et al. 2011; Kajino et al. 2019). The more popular models include core collapse SNe (e.g., Woosley et al. 1994) and neutron star mergers (e.g., Côté et al. 2018; Watson et al. 2019).

Y, Zr, Ba, La, Ce, Nd, and Eu are neutron-capture elements detectable in most of our sample stars across the observed wavelength. According to the recent work of Kobayashi et al. (2020), Y, Zr, Ba, La, Ce, and Nd are mostly produced through the s-process in the current universe, while Eu is produced through the r-process. The neutron-capture element abundances of our N-rich field stars are compared with MW stars in Figure 6. We see that the scatter of abundances in MW stars increases as metallicity decreases, given that feature lines are weaker in more metal-poor stars. Generally, we see consistency between our N-rich field stars and MW stars, but several stars with enhanced [Ba/Fe], [La/Fe], and [Eu/Fe] are also noticed.

Zoom In Zoom Out Reset image size

Two nuclear reactions are the major neutron excess sources in AGB stars: ¹³C(α, n)¹⁶O and ²²Ne(α, n)²⁵Mg. The first reaction dominates the low mass AGB stars, while the latter one is mainly found in massive AGB stars (Cristallo et al. 2015). As the number of free neutrons per iron seed increases, the s-process flow first seeds the light s-process peak (Sr−Y−Zr), extending to ¹³⁶Ba, and then reaches the heavy s-process peak (Ba−La−Ce−Pr−Nd), extending to ²⁰⁴Pb − ²⁰⁷Pb (Bisterzo et al. 2014). Therefore, the heavy s-process element to light s-process element ratio ([hs/ls]) is closely related to metallicity and initial stellar mass. In this work, [hs/ls] is defined as [hs/Fe] − [ls/Fe], where [hs/Fe] = ([Ba/Fe] + [La/Fe] + [Nd/Fe])/3, and [ls/Fe] = ([Y/Fe] + [Zr/Fe])/2. In our sample, there are six stars with all five s-process elements (Figure 7). Five stars show consistent [hs/ls] values as other MW stars in the same metallicity range. The grid lines of Cristallo et al. (2015) indicate these stars are enriched by AGB stars of masses around 3–5 M_⊙.

Zoom In Zoom Out Reset image size

The ratio between s-process element and r-process element is an indicator of the neutron-capture speed compared to β-decay, which could be related to the contribution of AGB stars (thus SNe Ia) over SNe II and neutron star mergers. Ba is commonly used to represent the heavy s-process elements, while Y for the light s-process elements. A gradual rise in [Y/Eu] and [Ba/Eu] with increasing metallicity is seen in Figure 8, which was addressed in, e.g., McWilliam (1998). The upper panel of Figure 8 shows that MW field stars, MW GC stars, and dwarf galaxy stars overlap in the [Ba/Eu]–[Fe/H] space. In the lower panel, [Y/Eu] of dwarf galaxies from Shetrone et al. (2003) and Venn et al. (2012) are slightly smaller compared to MW field stars and MW GC stars at a similar metallicity range, but with substantial overlap (also see Figure 19 of Venn et al. 2012). Though most N-rich field stars show consistent abundances with other studies, we are not able to distinguish their in situ or extragalactic origin here. Bisterzo et al. (2014) suggested that [Ba/Eu] ∼ − 0.7 is the typical value for a pure r-process. Most N-rich field stars show [Ba/Eu] abundances higher than the pure r-process value (green solid line in Figure 8), indicating a heterogeneous mixture of r- and s-processes. Star #9 and star #47 show [Ba/Eu] abundances close to the pure r-process value, indicating a strong r-process contribution.

Zoom In Zoom Out Reset image size

Though our sample stars have been confirmed to be mostly N-rich using CN-CH features around 4000 Å Paper I, further chemical tagging with more elements requires high-resolution spectra. Using the MIKE and APOGEE spectra, we find that most N-rich field stars show consistent Na, O, Mg, Al, Si, and Ca with GC stars of similar metallicities. This chemical similarity supports the GC origin of these N-rich field stars. Two stars with strong Mg depletion are also found. Strong Mg depletion is usually found in very metal-poor ([Fe/H ] < − 2.0) GCs (e.g., M15, M92). Two N-rich field stars show strong Mg depletion at [Fe/H ] ∼ − 1.3 is somewhat puzzling. One possible scenario is that accreted materials from companion stars may change the Mg abundances of these two N-rich field stars dramatically (Fernández-Trincado et al. 2017).

Based on the Na abundances, we find that most N-rich field stars are located in the transition region between FG and SG stars, thus it is difficult to verify their SG origin (see Figure 1). On the other hand, the Mg–Al figure (Figure 2) shows that half the stars in our sample have [Al/Fe ] > 0.5, which is quite different from GC FG stars. Interestingly, Fernández-Trincado et al. (2020) identified 29 mildly metal-poor ([Fe/H ] < − 0.7) field stars with [Al/Fe ] > 0.5. They suggested that these stars were ejected into the bulge and inner halo from GCs formed in situ and/or GCs formed in dissolved dwarf galaxy progenitors.

As our understanding of the MW formation rapidly improves in the Gaia era, GCs are now believed to have formed in both MW and dwarf satellite galaxies, and to have later merged together as the current Galactic GCs (e.g., Massari et al. 2019; Myeong et al. 2019). GCs formed in situ and ex situ tend to show different chemical and dynamical signatures. If the N-rich field stars escaped from GCs, they should carry similar chemodynamical information as the host GCs. Paper II showed that our parent N-rich field sample (∼100) included both stars formed in situ and stars formed ex situ using kinematics. In this work, one interesting star (star #9, large circles in Figures 4 and 5) shows consistently low [Mg/Fe], [Si/Fe], [Ca/Fe], [Ti/Fe], [Sc/Fe], [V/Fe], and [Co/Fe] compared to MW stars at similar metallicities, which is strong evidence that this star was formed in dissolved dwarf galaxies. The relatively higher [Al/Fe] of star #9 compared to dwarf galaxy field stars indicates that it is possibly enriched in the GC environment, as Al-rich stars can be found in accreted GCs. On the other hand, stars with − 1.25 ≤ [Fe/H] ≤ − 0.95 show [α/Fe] and iron-peak abundances consistent with other MW field stars, indicating that they were formed in situ. Moreover, we checked their kinematic information in Paper II, and found that star #9 shows 〈E〉 and 〈Lz〉 consistent with Gaia–Sausage–Enceladus (GSE) stars (Figure 9), while stars with − 1.25 ≤ [Fe/H] ≤ − 0.95 show in situ disk or halo kinematics (e.g., Naidu et al. 2020). Interestingly, star #9 shows [Ba/Eu] close to the r-process limit, which is consistent with other GSE stars observed in Aguado et al. (2021). This is a straightforward example to demonstrate that chemical tagging is possible to help decipher the origin of field stars. We also acknowledge the difficulties in judging the origin of field stars based on chemical and dynamical observational evidence, which is not always consistent. A more sophisticated probability estimation of their origin is left for further studies.

Zoom In Zoom Out Reset image size