Extending the Chemical Reach of the H3 Survey: Detailed Abundances of the Dwarf-galaxy Stellar Stream Wukong/LMS-1

We present the first detailed chemical-abundance analysis of stars from the dwarf-galaxy stellar stream Wukong/LMS-1 covering a wide metallicity range ($-3.5<\rm[Fe/H] \lesssim -1.3$). We find abundance patterns that are effectively indistinguishable from the bulk of Indus and Jhelum, a pair of smaller stellar streams proposed to be dynamically associated with Wukong/LMS-1. We confirmed a carbon-enhanced metal-poor star ($\rm[C/Fe]>+0.7$ and $\rm[Fe/H] \sim -2.9$) in Wukong/LMS-1 with strong enhancements in Sr, Y, and Zr, which is peculiar given its solar-level [Ba/Fe]. Wukong/LMS-1 stars have high abundances of $\alpha$ elements up to $\rm[Fe/H] \gtrsim -2$, which is expected for relatively massive dwarfs. Towards the high-metallicity end, Wukong/LMS-1 becomes $\alpha$-poor, revealing that it probably experienced fairly standard chemical evolution. We identified a pair of N- and Na-rich stars in Wukong/LMS-1, reminiscent of multiple populations in globular clusters. This indicates that this dwarf galaxy contained at least one globular cluster that was completely disrupted in addition to two intact ones previously known to be associated with Wukong/LMS-1, which is possibly connected to similar evidence found in Indus. From these $\geq$3 globular clusters, we estimate the total mass of Wukong/LMS-1 to be ${\approx}10^{10} M_\odot$, representing ${\sim}1$% of the present-day Milky Way. Finally, the [Eu/Mg] ratio in Wukong/LMS-1 continuously increases with metallicity, making this the first example of a dwarf galaxy where the production of $r$-process elements is clearly dominated by delayed sources, presumably neutron-star mergers.

Because the evolution of these dwarf galaxy streams and substructures was interrupted due to the shutdown of their star formation at the moment of their accretion, they provide a unique local window into the properties of galaxies that existed at high redshift (Boylan-Kolchin et al. 2015, 2016).Also, as these systems sample a wide range of stellar masses (10 6 ≲  ★ / ⊙ < 10 9 ; see Naidu et al. 2022 andSharpe et al. 2022), they provide a laboratory for us to test the universality of galaxy formation/evolution processes as we can compare them with observations of intact dwarfs in the Local Group (Tolstoy et al. 2009 andSimon 2019 for reviews).Conveniently, this approach comes with the advantage that member stars of these disrupted dwarfs are typically much closer and, hence, brighter than their counterparts located in surviving Milky Way satellites.This makes them more easily accessible to high-resolution (R ≥ 20,000) spectroscopy, from which detailed chemical abundances can be ob-tained for significant amounts of stars, allowing us to understand the properties of their progenitor systems.This "near-field" approach to galaxy evolution at the smallest scales is even more appealing given that not even current and future facilities (such as JWST or 20-30 m ground telescopes) will be able to spatially resolve such low-mass galaxies at the redshift range (0.5 <  ≤ 2.0) probed by these halo debris (Myeong et al. 2018(Myeong et al. , 2019;;Koppelman et al. 2019a;Forbes 2020;Kruĳssen et al. 2020;Naidu et al. 2020;Callingham et al. 2022, see also Naidu et al. 2021 for a relevant discussion).
In this contribution, we present the first detailed chemical abundance analysis of Wukong/LMS-1 (Naidu et al. 2020;Yuan et al. 2020;Malhan et al. 2021).This substructure was identified by Yuan et al. (2020) who named it the "low-mass stellar-debris stream 1" (LMS-1).Independently, Naidu et al. (2020) identified "Wukong" 2 , a prominent group of stars in integrals-of-motion space apparently dissociated from any previously known disrupted dwarf.The connection between LMS-1 and Wukong was quickly recognized for their indistinguishable dynamics, including the association with at least two globular clusters, NGC 5024 (M53) and NGC 5053, and two other stellar streams, Indus and Jhelum (Shipp et al. 2018;Bonaca et al. 2019Bonaca et al. , 2021;;Malhan et al. 2022).Out of the known disrupted dwarf galaxies, some listed above, Wukong/LMS-1 is especially interesting due to its predicted relatively high stellar mass of ∼10 7  ⊙ (Malhan et al. 2021), which is similar to classical Milky Way satellites such as Sculptor and Fornax dSph galaxies (McConnachie 2012).Therefore, our goal is to constrain Wukong/LMS-1's star-formation history, its production of the heaviest elements via prompt and/or delayed sources of neutron-capture processes, and even look for signatures of dissolved globular clusters in it.Then, we put our results in context by comparing with known Milky Way satellite galaxies of similar mass as well as chemical-evolution models.
1 Although we list Sequoia as an independent disrupted dwarf, we recognize the current dispute in the literature regarding whether or not this population could simply be part of the more massive GSE (Koppelman et al. 2020;Amarante et al. 2022;Limberg et al. 2022;Horta et al. 2023). 2 Named after Sun Wukong, the celestial Monkey King from Journey to the West.See Naidu et al. (2020) for the complete rationale.This work is organized as follows.Section 2 includes all things related to our observations, data reduction, and radial velocity (RV) measurements.Our methodology for obtaining stellar parameters and abundances from high-resolution spectra is described in Section 3. Section 4 is reserved for the presentation of our results.In Section 5, we provide our concluding remarks and a brief discussion.

Observations
We observed 13 Wukong/LMS-1 targets (Wuk_1 to _13)3 , which were selected as best-suited for covering a wide metallicity range within the telescope time allocation available.These were originally identified by Naidu et al. (2020) using data from the Hectochelle in the Halo at High Resolution (H3) survey (Conroy et al. 2019), including orbital energy and angular momentum criteria, but also an [Fe/H] < −1.45 cut to avoid contamination by GSE stars.The observed sample is also contained in the "high signal-to-noise" (high-/) sample of Wukong/LMS-1 stars from Johnson et al. (2022), also with H3 data, who refined Naidu et al.'s (2020) et al. (2020), which was developed independently from the H3 survey papers.The full criteria is written below.Because GSE-like mergers are not expected to deposit debris on such acute polar orbits (see Amarante et al. 2022, Limberg et al. 2023 for discussion), we can be confident that our APOGEE metal-rich target is a genuine member of Wukong/LMS-1.
In Figure 1, we exhibit all our targets in projected action space within the McMillan (2017) Galactic model potential.The action vector is written as J = (  ,   ,   ), where   ,   , and   are the radial, azimuthal, and vertical components in a cylindrical frame.Both of the above-mentioned Wukong/LMS-1 samples and the globular clusters NGC 5024 and NGC 5053 are also plotted.For these orbit calculations, we integrated for 20 Gyr forward using the AGAMA library (Vasiliev 2019).Positions and proper motions on the sky are from Gaia DR3.RVs were determined by us for the stars we observed (see below).Other RVs employed were taken from their parent samples.Distances come from spectro-photometric fits, MINESweeper (Cargile et al. 2020) for H3 and StarHorse (Queiroz et al. 2020(Queiroz et al. , 2023) ) for APOGEE.For NGC 5024 and NGC 5053, all phase-space information is from Vasiliev & Baumgardt (2021).We realized each orbit 100 times in a Monte Carlo scheme assuming Gaussian uncertainties for these quantities.The final adopted values are the medians of the resulting distributions.The adopted distance from the Sun to the Galactic center is 8.2 kpc (Bland-Hawthorn & Gerhard 2016), the circular velocity at this position is 232.8 km s −1 (McMillan 2017), and the assumed peculiar motion of the Sun is (, , ) ⊙ = (11.10,12.24, 7.25) km s −1 (Schönrich et al. 2010).
We note that the model potential as well as Galactic fundamental parameters adopted here are different from previous H3 survey works (Naidu et al. 2020;Johnson et al. 2022).Therefore, the criteria used by these authors to select Wukong/LMS-1 stars become slightly different after our above-described calculations.In the spirit of making this paper self-sufficient, we provide updated values for the kinematic/dynamical quantities that define the Wukong/LMS-1 structure below.Nevertheless, we reinforce that the original target selection was made simply based on the H3 survey samples from Naidu et al.
(2020) and Johnson et al. (2022) plus covering the largest possible metallicity range within the available telescope allocation.
Following Johnson et al. (2022), we have: where  total =   + |  | +   (Figure 1) and  = arccos (  /) refers to the orbital inclination, which characterizes the direction of the angular momentum vector L = (  ,   ,   ) in a Galactic Cartesian frame.Within these definitions,  = is the total angular momentum and   ≡   is the vertical component of it.These criteria are accompanied by those from Naidu et al. (2020): • −1000 <   /(kpc km s −1 ) < 0 and where  is the total orbital energy.Apart from these cuts, Naidu et al. (2020) also removed possible Sagittarius stream interlopers using the simple method of Johnson et al. (2020), which has been shown to be likely complete (Peñarrubia & Petersen 2021).Hence, when selecting for Wukong/LMS-1, one can use fully eliminating Sagittarius contamination within its   range.Naidu et al. (2020) also included a cut in orbital eccentricity to eliminate GSE stars.However, this is redundant with the action-space selection of Johnson et al. (2022) that requires   >   .
We observed all the Wukong/LMS-1 stars with the Magellan Inamori Kyocera Echelle (MIKE; Bernstein et al. 2003) spectrograph installed in the Magellan Clay telescope (6.5 m) located at Las Campanas Observatory, Chile.For all stars over all observing runs (June 2022 for H3 follow-up and May 2023 for APOGEE), we used 0.7 ′′ slit and 2×2 on-chip binning.This configuration leads to resolving powers of R ∼ 35,000 and 28,000 for the blue (wavelength  < 5000 Å) and red ( > 5000 Å) arms of MIKE spectra, respectively.All data were reduced using the CarPy 5 package (Kelson 2003).The final 4 We call the attention to the difference in definition of  total between ours and Naidu et al.'s (2020) work, also Johnson et al. (2022).These authors assumed the vectorial definition  total = √︃  2  +  2  +  2  . 5https://code.obs.carnegiescience.edu/mike.(Barklem et al. 2005;Marino et al. 2008;Kirby et al. 2009;Cohen et al. 2013;Roederer et al. 2014;Jacobson et al. 2015, see Ji et al. 2023a).The black line was used to determine the  t for the most metalpoor star analysed (Wuk_6; see text).Wuk_4 is shown with red edges in both panels (see Section 4.1).

Radial velocities and spectra normalization
We derived RVs for all Wukong/LMS-1 stars by cross-correlating against a high-/ MIKE spectrum of the metal-poor standard HD 122563 using the Labeling Echelle Spectra with SMHR and Payne (LESSPayne6 ; A. P. Ji, in preparation) code.In a nutshell, LESSPayne combines Spectroscopy Made Harder (smhr7 ; Casey 2014) with Payne4MIKE8 (Ting et al. 2019) 9 and consolidates it into a single package.As of now, the RV measurement routine within LESSPayne is a carbon-copy of smhr's.The formal statistical uncertainty of RVs from MIKE spectra ("RV MIKE " in the top panel of Fig- ure 2 and Table 1) could, in principle, reach ∼0.1 km s −1 .However, a systematic error of ∼1 km s −1 is introduced due to slit centering and wavelength calibration (see discussion by Ji et al. 2020b).For the purpose of this work, we are satisfied that the membership of our stars does not depend on the choice of RV value, either MIKE or H3/APOGEE.Indeed, at this RV precision, distances are the dominant source of errors for orbital parameters.The maximum difference between our MIKE RVs and H3 ones is ≈10 km s −1 (Wuk_4), but is usually <2 km s −1 (Figure 2).See Section 4.1 for the possibility that Wuk_4 is in a binary system.For the APOGEE star Wuk_14, the difference in RV is 0.5 km s −1 .
With the reduced data and RVs at hand, we proceed to stitch orders and normalize the MIKE spectra using cubic spline functions.LESSPayne initializes smhr-like files from the best-fitting Payne4MIKE synthetic spectrum.This procedure drastically accelerates the normalization and equivalent width measurements, as it identifies where absorption features occur and masks them.This method is identical to the one described in Ji et al. (2023b), with the difference that, now, the whole process has been packaged into LESSPayne.The entire spectrum was inspected for all stars, but the LESSPayne continuum needed minor fixes only at the bluest orders, where / is lower, or when prominent wide absorption was present, such as Ca ii K/H (3900-4000 Å) and/or the C-H G band (∼4300 Å).

METHODS
For this project, we considered the atomic and molecular line list from Ji et al. (2020a), which was adapted from the linemake10 compilation (Placco et al. 2021).This line list was also used by Ji et al. (2020a) to analyze the Indus and Jhelum streams that could be associated with Wukong/LMS-1 and which we compare to in Section 4. We rejected all lines at <3860 Å in our analysis due to the lower / at the bluest portions of the MIKE spectra.For our equivalent width measurements, we assumed Gaussian profiles.Lines that required Voigt profiles were rejected, with the exception of those in the Mg i triplet (5150-5200 Å).Abundances for Mg from these strong lines are consistent with others considered.Overall, we fitted between 70 (for our most metal-poor star) and 380 (most metal-rich) lines in each MIKE spectrum.Individual line measurements are provided in supplementary online tables.

Stellar Parameters
We computed fully spectroscopic stellar parameters for all stars under the assumption of local thermodynamic equilibrium (LTE) using LESSPayne, which wraps the radiative transfer code MOOG Lastly, we recalibrated the  eff for each star to the "photometric scale" of Frebel et al. (2013) and, then, rederived log  and  t by repeating the same steps described above.We reanalysed from scratch a couple of stars from Ji et al. (2020a) with our method and verified that these authors' photometric stellar parameters are compatible (1) with those revised with the Frebel et al. ( 2013) correction.On average,  eff values become ≈200 K hotter after this step.For log  and [M/H], the corrected values are, on average, 0.67 dex and 0.24 dex larger, respectively.For  t , the final values are 0.07 km s −1 lower on average.Note that we do not just apply constant offsets, but rather the stellar parameters of each star change independently according to their recalibrated  eff and these listed values are simply the average corrections.In Table 1 (also all figures), we provide only corrected parameters.Stellar parameters prior to recalibration are provided as online supplementary material.Throughout the remainder of this work, we consider only recalibrated stellar parameters, including for the abundance analysis (Section 3.2).
Apart from just statistical uncertainties, which come from the fitting process of individual lines as well as from the slopes of the excitation/ionization balance fits, it is crucial to consider systematics that can affect our stellar parameters and which are propagated into abundances.Systematic errors in stellar spectroscopic analysis come, in principle, from departures from the 1D LTE assumption, i.e., 3D and/or NLTE effects, which are especially relevant for metalpoor giants (Asplund 2005).In order to obtain empirically motivated systematic errors, we looked at the sample of low-metallicity stars analysed by Ezzeddine et al. (2020), which spans a wide range of stellar parameters that completely encompasses our sample; 4600 <  eff /K ≲ 5300, 1 < log  ≲ 3, −3.50 ≲ [Fe/H] < −1.25, and 1 <  t /(km s −1 ) ≲ 2. These authors provided different sets of stellar parameters for each of their stars, including 1D LTE with the Frebel et al. ( 2013) correction and 1D NLTE.We verified that the average differences between these stellar parameter estimates are 100 K, 0.5 dex, 0.1 km s −1 , and 0.2 dex for  eff , log ,  t , and [Fe/H] and these values are adopted as systematic errors, summing in quadrature with the statistical ones to obtain final uncertainties for these quantities.For Wuk_6, we adopted a  t statistical uncertainty of 0.12 km s −1 , which is the intrinsic scatter from the quadratic fit to the log  versus  t relation.We note that, despite the apparently excessively large log  errors,  eff continues to have the most important impact on abundance uncertainties.
A comparison between our MIKE parameters and those from the H3 survey/MINESweeper is provided in Figure 3.For  eff and log , these are compatible (1) for all stars.For [Fe/H], H3 values are larger by ∼0.3 dex, but this level of systematics can be attributed to NLTE effects (see figure 2 of Ezzeddine et al. 2020).In comparison to APOGEE parameters, we found  eff and log  differences of ∼300 K and ∼0.7 dex, respectively, but with equivalent [Fe/H] values (also Figure 3).We note, however, that, before applying the Frebel et al. ( 2013) photometric correction, our purely spectroscopic  eff and log  would be compatible (1) with APOGEE values.With respect to the H3 values, this survey adopts spectro-photometric fits to obtain stellar parameters from broad-band photometry and model isochrones.Henceforth, it is not unexpected that our photometrically recalibrated  eff and log  are similar, as is the case between ours and Ji et al.'s (2020a) photometric stellar parameters.

Abundances
We obtained abundances for up to 29 species of 24 elements.Equivalent widths were used for the abundances of Na i, Mg i, Si i, K i, Ca i, Ti i, Ti ii, Cr i, Cr ii, Ni i, Zn i, and Sr i, as well Fe i and Fe ii.For heavily blended features, molecular features, or those requiring hyperfine splitting, we defaulted to spectral synthesis.For the task, LESSPayne performs a  2 minimization over local continuum with RV and abundance as free parameters, as well as a smoothing parameter to account for resolution.This approach was employed for C-H and C-N molecules, Al i, Sc ii, V i, V ii, Mn i, Co i, Sr ii, Y ii, Zr ii, Ba ii, La ii, Eu ii, and Dy ii.We also synthesized the Si i line at 3905Å due to substantial blending.Additionally, we applied this technique to derive 5 upper limits for O i, as well as other elements, for all our stars.For C-H and C-N, we adopted 12 C/ 13 C = 9.For Ba and Eu, solar -process isotopic ratios were assumed (Sneden et al. 2008).For [Ti/Fe], [V/Fe], [Cr/Fe], and [Sr/Fe] ratios, we adopt Ti ii, V i, Cr i, and Sr ii as our fiducial values (Figure 4).
Final abundances were calculated as inverse-variance weighted averages, taking into account uncertainties from individual lines for each species.Our adopted procedure is similar to Ji et al. (2020b), which neglects correlations between stellar parameters (McWilliam et al. 2013).The one difference between our methodology and these authors' is the inclusion of an error floor on a line-by-line basis ( sys ), which should encapsulate additional unknowns coming from, for instance, atomic data or the 1D model atmospheres (see Ji et al. 2020a).For this purpose, we adopted a constant  sys = 0.1 dex floor for most lines.For some species with hyperfine splitting, where the smoothing kernel might be degenerate with abundance, such as Sc, Mn, and Ba, we add an extra 0.1 dex, i.e.,  sys = 0.2 dex for all lines.For the C-N molecular band, which is located in the bluest part of MIKE spectra (3865-3885 Å) where the / is low and the continuum placement is difficult, we implemented  sys = 0.3 dex.The Al line at 3961 Å, which is the only one considered for this element, not only suffers from the same caveats as the C-N, but is also at the wing of a hydrogen line (H), so we employed  sys = 0.3 dex for it as well.We provide the relevant equations below, which are analogous to equations 1 to 5 in Ji et al. (2020b), but incorporating our modification and with updated terminology.
For a certain line  of a given element/specie X, its total uncertainty (  ) of the associated abundance (  ) can be written as the quadrature sum of the different sources of error, i.e., where  ,stat is the statistical error, which comes from spectrum noise and the line-fitting procedure, and  SP is the total stellar-parameter error budget; In this equation,   1.Because these quantities retain their sign, we refer to them as " ,••• ".Finally, the inverse-variance weights can be assigned to each line, and the final abundance can be computed as where  is the total amount of lines available for X.Uncertainties in (X), as well as abundance ratios, are propagated in an identical fashion to Ji et al. (2020b).Again, to make this paper self-sufficient, we reproduce these authors' equations 6 through 10 (see their appendix) below.
where this total statistical uncertainty for abundance (X) includes the weighted standard error across abundances   for  different lines  of element/specie X as well as spectrum noise.Note that when  = 1, the first term of Equation 5 goes to zero and, in that situation, the second term is fully responsible for propagating line-byline errors (including our  sys ) to the final  stat through the weights pre-computed by Equation 3.For stellar parameters, We recall that covariance between stellar parameters ( eff , log ,   , and [M/H]) is neglected (see McWilliam et al. 2013).Then, the total error budget for an abundance ratio [X/H] between element/specie X and hydrogen is obtained by summing statistical and stellar-parameter uncertainties in quadrature: Finally, the uncertainty of an abundance ratio between elements/species X and Y accounts for covariance between X and Y through stellar parameters: Our abundance inventory is shown in Figure 4. Our Wukong/LMS-1 stars (yellow symbols) are plotted against a compilation of metalpoor stars (grey dots) from the Stellar Abundances for Galactic Archaeology (SAGA) database (Suda et al. 2008(Suda et al. , 2017)), including the works of Fulbright (2000), Barklem et al. (2005) 2022).All abundance information, both ours and SAGA's, was normalized to the solar composition of Asplund et al. (2009).Whenever we mention the ratio between an element X and Fe ([X/Fe]), we adopt Fe i as reference.We highlight the red dashed line in the upper left panel of Figure 4, which delineates the boundary for "carbon-enhanced metal-poor" stars (CEMP; [C/Fe] > +0.7 and [Fe/H] < −1; Beers & Christlieb 2005, Aoki et al. 2007, Frebel & Norris 2015, Frebel 2018).Dotted lines portray the solar-level abundances ([X/Fe] = 0) in all panels.
For Indus and Jhelum stellar streams, we use the line measurements from Ji et al. (2020a), but we recomputed the average abundance ratios using our described methodology (Section 3.2) rather than the weights from their paper.We proceed into this section with the assumption that our analysed abundances for Wukong/LMS-1, Indus, and Jhelum are on a consistent scale and with the understanding that these stellar streams are all associated through a common progenitor dwarf galaxy.If these hypotheses hold true, the bulk of their chemical-abundance patterns should be effectively indistinguishable from each other.Having said that, we caution that, although chemistry is a great tool for falsifying associations between substructures, similar abundance patterns do not automatically confirm common origins, especially if these are not too different from the underlying halo population.

A CEMP star with peculiar neutron-capture signatures
It has been known for at least a couple of decades that the fraction of CEMP stars increase as function of decreasing metallicity (Beers et al. 1992;Norris et al. 1997;Rossi et al. 1999Rossi et al. , 2005;;Lucatello et al. 2006;Lee et al. 2013;Placco et al. 2014Placco et al. , 2018;;Yoon et al. 2018, andsee Arentsen et al. 2022 for a compilation).The evolution of the CEMP fraction with [Fe/H] appears to be identical between the Milky Way's halo and UFD galaxies (Ji et al. 2020b).However, detailed studies for some more massive dwarfs, such as classical dSph ones, show that some discrepancy might exist with the Milky Way, in particular for Sculptor (Skúladóttir et al. 2015(Skúladóttir et al. , 2021(Skúladóttir et al. , 2023;;Chiti et al. 2018) and, possibly, Sagittarius (Chiti & Frebel 2019;Chiti et al. 2020, but see Limberg et al. 2023).
In this context, we confirm that Wuk_4, one of our most metalpoor Wukong/LMS-1 members ([Fe/H] = −2.89±0.19), is a CEMP star ([C/Fe] = +0.74± 0.25; top left panel of Figure 4, symbol with red edge).The carbon-abundance correction for evolutionary effects is only +0.01 dex (Placco et al. 2014).After, applying the carbon correction to all other stars in our Wukong/LMS-1 sample, none of them turned out to be CEMP.The corrected [C/Fe] values for all stars are provided alongside the full abundance table as supplementary material.Wuk_4 had already been identified as a CEMP candidate by Lucey et al. (2023) through spectro-photometric data from Gaia DR3.Wuk_4 also shows strong enhancement in Sr, Y, and Zr (+0.5 ≲ [X/Fe] ≤ +1.0; upper panels of Figure 5).Interestingly, this is not accompanied by significant enrichment in either Ba ([Ba/Fe] = +0.12± 0.24) or La (only upper limit found) as would be expected if Wuk_4 was a "typical" CEMP- star (Beers & Christlieb 2005;Frebel 2018), i.e., a CEMP star also enhanced in the slow (-) neutron capture process (defined as [Ba/Fe] > +1.0).Apart from Wuk_4, other Wukong/LMS-1 stars have effectively identical chemical compositions to the bulk of Indus and Jhelum stars in all abundance panels in Figure 5 (Sr, Y, Zr, Ba, La).
In Figure 5, we exhibit the same SAGA data base low-metallicity compilation from Figure 4, but, now, dividing into carbon-normal (grey points), regular CEMP (orange diamonds), and CEMP- (blue triangles) stars.From the top panels, it becomes clear that CEMP- stars are also preferentially enhanced in Sr, Y, and Zr.Nevertheless, there are a few examples of ordinary CEMP stars with high values of [Sr/Fe], [Y/Fe], and/or [Zr/Fe] similar to Wuk_4.In any case, it is also rather suspicious that Wuk_4 is the only star with significant (≈10 km s −1 , ≳5) RV difference between H3's and our measurement (Section 2.2), possibly indicating a binary system.Indeed, binarity is believed to be the conventional pathway for the formation of CEMP- stars as they would experience mass transfer from an asymptotic giant branch companion (e.g., Lucatello et al. 2005 andHansen et al. 2016 for RV monitoring studies and Travaglio et al. 2004 for nucleosynthesis).Nevertheless, CEMP-no stars, those without any signatures of -process enhancement ([Ba/Fe] < 0.0) are also, sometimes, found in long period binaries (Arentsen et al. 2019;Bonifacio et al. 2020).Deciphering whether or not mass transfer in a binary can create the excess of Sr, Y, and Zr in Wuk_4 without high Ba will demand a systematic investigation of other CEMP stars with similar abundance patterns in the future.The marker with red edge represents the CEMP star in our sample (Wuk_4).The yellow square is the most metal-rich star followed-up from APOGEE (Wuk_14).Pink and purple crosses with white edges are stars from Indus and Jhelum stellar streams, respectively.In the left panel, blue (dotted), orange (dashed), and green (solid) lines are chemical evolution trajectories for galaxies with stellar masses  ★ / ⊙ of 10 6 , 10 7 , and 10 8 , respectively (Wanajo et al. 2021, see text).In the middle and right panels, the black line shows the median abundances from the SAGA data base compilation of metal-poor stars in bins of 0.

𝛼 elements reveal a relatively massive dwarf galaxy
Now, we look at the [/Fe] ratios in Wukong/LMS-1 stars, specifically the abundances of Mg and Ca.We only derived upper limits for O and Si is less reliable than Mg and Ca in the low-metallicity regime (see discussion in Ji et al. 2020a).The behaviour of  abundances is a great tracer of the overall star-formation history of a galaxy (Tinsley 1979; Matteucci 2012).More massive systems are expected to enrich themselves (i.e., reach higher metallicities) via core-collapse supernovae before the occurrence of Type Ia supernovae, which is delayed by ≳100 Myr (e.g., Maoz & Mannucci 2012;de los Reyes et al. 2022).Because core-collapse supernovae produces mostly  elements while the Type Ia create almost exclusively iron-peak elements (for a review, see Nomoto et al. 2013), a characteristic downturn in [/Fe] is expected at a certain metallicity (referred to as the "knee"; Matteucci & Greggio 1986).The location of this knee in [Fe/H] depends on the star-formation and outflow efficiency, hence the mass, of a given galaxy (e.g., Matteucci &Brocato 1990 andTolstoy et al. 2009).This is illustrated in Figure 6 (left panel) by the simulated chemical-evolution trajectories of Wanajo et al. (2021, their figure 4, case 1) for dwarf galaxies with different stellar masses (10 6 ≤  ★ / ⊙ ≤ 10 8 ) at redshift  = 0, similar to canonical dSph Milky Way satellites and encompassing the expected value for Wukong/LMS-1's progenitor.We note that we subtracted 0.25 dex from their model's [Mg/H] values so the "plateau" is positioned at about +0.4 dex.These recalibrated versions of the models will also be used in Section 4.4.
In Wukong/LMS-1, our derived abundances of both Mg and Ca reveal that the [/Fe] ratio remains high (∼0.3-0.4 dex) up to [Fe/H] ≳ −2 (Figure 6).This is, indeed, similar to somewhat massive surviving Milky Way satellites (Kirby et al. 2011;Reichert et al. 2020).Before our analysis, the evidence for Wukong/LMS-1's progenitor being a relatively massive dwarf ( ★ ∼ 10 7  ⊙ ) came from tentative dynamical -body modeling (Malhan et al. 2021), the scaling relation between the total mass of its globular clusters and a galaxy's mass (also Malhan et al. 2021), or simple star counts (Naidu et al. 2022).Hence, we provide the first evidence in favor of this hy-pothesis from chemistry.An identical -element abundance pattern is seen in Indus and Jhelum (pink and purple crosses, respectively, in Figures 5 and beyond; Ji et al. 2020a).As previously mentioned, although this chemical similarity between these streams does not fully confirm their association, it still corroborates such scenario.
Although the models in Figure 6 exemplify how a galaxy's mass correlates with its chemical evolution, we do not claim to be actually measuring the stellar mass of Wukong/LMS-1 from abundances.Systematics make the uncertainties of our [Mg/Fe] and [Ca/Fe] estimates to be typically 0.2 dex.For reference, the median statistical uncertainties are ≈0.1 dex for these [/Fe] ratios (top right corner in middle and right panels of Figure 6).Not only this precision does not allow us to differentiate between those models, but the models themselves carry potentially even worse systematics such as supernovae and/or NSM yields.Nevertheless, we reinforce that the constant, within errors, [/Fe] up to [Fe/H] ∼ −2 is in conformity with the behaviour seen in the data of nearby massive ( ★ / ⊙ ≥ 10 6 ) dwarfs.
It is also relevant that the most metal-rich star ([Fe/H] = −1.26± 0.20, Wuk_14; yellow square in all figures) in the sample has lower [Mg/Fe] and [Ca/Fe] than the bulk of our observed Wukong/LMS-1 members ([Fe/H] ≤ −1.75) by 0.1-0.2dex (middle and right panels of Figure 6).This is, perhaps, not a particularly surprising result given that we observed this additional star specifically with the goal of testing the standard chemical enrichment scenario as discussed above.Nevertheless, we notice that Jhelum also contains a low- star at [Fe/H] = −1.45,which adds to the emerging picture where Wukong/LMS-1 progenitor experienced quite a simple chemical evolution pathway.It is also informative that Wukong/LMS-1 shows no evidence for additional bursts of star formation, which would cause the [/Fe] to actually increase at higher [Fe/H], a phenomenon that happens to some dwarfs, such as Sagittarius, Fornax, and the Magellanic Clouds (see Nidever et al. 2020 andHasselquist et al. 2021), and is expected to be caused by their interaction with their massive host (in this case, the Milky Way).We did not find any trustworthy Wukong/LMS-1 candidates at metallicities higher than Wuk_14 neither in H3 nor in APOGEE.

A pair of stars with abundances reminiscent of multiple stellar populations in globular clusters
Almost all globular clusters are known to contain more than just a single, sometimes several, well-defined sequences in optical/ultraviolet color-magnitude diagrams (Piotto et al. 2015;Milone et al. 2017).Color differences between these distinct tracks are caused by star-to-star variations in their abundances of light elements, including correlated enhancements in N and Na (e.g., Gratton et al. 2004).Such phenomenon is referred to as "multiple stellar popula-tions" (Bastian & Lardo 2018;Milone & Marino 2022).Star clusters with multiple stellar populations have been detected not only in the Milky Way, but also other galaxies in the Local Group, including the Magellanic Clouds (Mucciarelli et al. 2009;Dalessandro et al. 2016) and even Fornax dSph (Larsen et al. 2014).Most recently, evidence has been provided for multiple stellar populations in globular-cluster stellar streams (Balbinot et al. 2022;Martin et al. 2022;Usman et al. 2024).Interestingly, a clear detection of a N-/Na-rich star was actually first made in Indus (Ji et al. 2020a), which holds even after NLTE corrections to Na (Hansen et al. 2021).
Wukong/LMS-1 contains an additional pair of N-/Na-rich stars (top panel of Figure 7), reinforcing the idea that its progenitor dwarf galaxy had not only NGC 5024 and NGC 5053 as globular clusters, but also at least one other that has been fully disrupted (Hansen et al. 2021), which is not unexpected for its mass (Eadie et al. 2022).In order to guarantee that this result is robust against NLTE effects, we performed corrections following Lind et al. (2011).In practice, we corrected Na abundances line by line and, then, recomputed the weighted averages using our formalism (Section 3.2).In this process, uncertainties and weights remain unchanged.For consistency, we applied the same approach to both Indus and Jhelum stars.As can be appreciated from Figure 7, NLTE corrections to Na exacerbate the differences between the bulk of our Wukong/LMS-1 members and its enriched stars; original LTE abundances are plotted in the background as transparent symbols.We note that both NGC 5024 and NGC 5053 are also very metal-poor ([Fe/H] ≲ −2; Kruĳssen et al. 2019 and references therein) and contain multiple stellar populations (see their ultraviolet color-magnitude diagrams in Piotto et al. 2015) From the bottom panel of Figure 7, it is intriguing that both of our N-/Na-rich Wukong/LMS-1 stars (Wuk_5 at [Fe/H] = −2.42± 0.25 and Wuk_11 at [Fe/H] = −2.37±0.20)have metallicities compatible (1) to their analog in Indus (Indus_0; [Fe/H] = −2.32 ± 0.22).We speculate that this could indicate that all these stars originated from a single disrupted globular cluster, though there is still not enough evidence to make a clear association.Nevertheless, if more stars at this metallicity of ≈−2.4 dex are found to have excess of N and Na in Indus, Jhelum, and the large Wukong/LMS-1, the scenario where they all belonged to the same, now completely disrupted, globular cluster would be reinforced.Furthermore, this underscores the exciting possibility of identifying more enriched globular-cluster stars embedded within other dwarf-galaxy streams, which would be further evidence in favor of the multiple stellar populations phenomenon being ubiquitous across different environments.Now that we have strong indication that Wukong/LMS-1 had ≥3 globular clusters, we can utilize this information as independent constraint on the total mass of this dwarf (e.g., Forbes 2020 andCallingham et al. 2022).We employ the relationship between the total number of globular clusters in a galaxy and its halo virial mass from Burkert & Forbes (2020, their equation 1).This exercise gives a total mass of ≈10 10  ⊙ for Wukong/LMS-1.This value, indeed, translates into a stellar mass of ∼10 7 assuming the stellar-to-halo mass relation from Rodríguez-Puebla et al. (2017) at redshift  = 1 (∼8 Gyr ago in Planck Collaboration et al. 2020 cosmology), consistent with expectations for Wukong/LMS-1 (Malhan et al. 2021;Naidu et al. 2022).From this line of reasoning, Wukong/LMS-1 might have contributed with ∼1% of the present-day total mass of the Milky Way (e.g., Bland-Hawthorn & Gerhard 2016).

The rise of the 𝑟-process in Wukong/LMS-1
Finally, we look at the abundance patterns of rapid (-) neutron capture process elements in Wukong/LMS-1.In 2017, the elec-tromagnetic counterpart of the neutron-star merger (NSM) event GW170817 (Abbott et al. 2017a,b,c) provided confirmation that this site is capable of producing copious amounts of heavy elements via the -process (Drout et al. 2017;Kilpatrick et al. 2017;Shappee et al. 2017).Notwithstanding, although NSMs could be the only site for the -process, evidence has been presented that additional sources are involved (see the review by Cowan et al. 2021), in particular a prompt source might be needed to explain the full abundance patterns of both the Milky Way (Côté et al. 2019;Haynes & Kobayashi 2019;Kobayashi et al. 2020;Tsujimoto 2021) and some of its dSph satellites (Skúladóttir et al. 2019;Reichert et al. 2020;Skúladóttir & Salvadori 2020).The timescales for -process enrichment embedded into the chemical abundances of stars hold clues to its dominant astrophysical site(s).
As can be seen in the top panel of Figure 8, the [Eu/Fe] ratio, where Eu is mostly produced by the -process (see Sneden et al. 2008), increases from <+0.1 dex at [Fe/H] ∼ −2.4 all the way up to >+0.7 dex at [Fe/H] ∼ −1.3.The [Ba/Eu] ratio of Wukong/LMS-1 stars is effectively constant at ∼−0.4 dex, confirming that the production of these elements is dominated by the -process.For reference, a "pure" -process signature is close to [Ba/Eu] = −0.8(e.g., Bisterzo et al. 2014).These abundances overlap with both Indus and Jhelum, although the data for these other stellar streams do not cover as large of a metallicity interval.We also recall the existence of the extremely -process rich star Indus_13 (Hansen et al. 2021), which is labeled in Figure 8.The previously mentioned Wanajo et al. (2021) chemical-evolution sequences (Section 4.2) are plotted in the background of Figure 8.These models contain -process production exclusively from NSMs.The current data for Wukong/LMS-1 seems compatible with this scenario, without the need for a prompt source (such as core-collapse supernovae) where [Eu/Fe] (or [Eu/Mg] as in the bottom panel of Figure 8) would be constant over the lowmetallicity regime.Note that we do not claim that NSMs are the only source of -process in Wukong, but rather the dominant one within the metallicity range probed by our current data.
An increasing trend of [Eu/Fe] with [Fe/H] has also been claimed to exist in UMi dSph (Cohen & Huang 2010).However, this behaviour is accompanied by a huge scatter, which suggests that UMi actually experiences stochastic -process enrichment, which, indeed, would be in line with its lower stellar mass ( ★ ∼ 10 5  ⊙ ) in comparison to Sculptor, Fornax, and Sagittarius dSph galaxies ( ★ > 10 6  ⊙ ).Similar observations have also been made in another dwarf-galaxy stellar stream, the so-called "Typhon" (Tenachi et al. 2022;Dodd et al. 2023, andJi et al. 2023b for abundances).Nevertheless, the available data for that stream covers a shorter metallicity interval and the differences in [Eu/Fe] are smaller than in Wukong/LMS-1.Also, in Typhon, this result depends on a single star at [Fe/H] < −2.Therefore, Wukong/LMS-1 constitutes the first example of a dwarf galaxy, though in the form of a stellar stream, with -process enrichment that is clearly dominated by delayed sources, presumably NSMs.Of course, additional measurements of Eu at lower metallicities ([Fe/H] < −2.5) will be necessary to test for the presence of a prompt -process source in Wukong/LMS-1.Unfortunately, all stars analysed in this work at such low-metallicity regime are relatively hot ( eff ≳ 5100 K; Table 1, Wuk_4, 6, and 7), which impedes us from obtaining detections of Eu with the lines considered (4129Å, 4205Å, 4435Å, 4522Å, and 6645Å).Wuk_4).The yellow square is the most metal-rich star followed-up from APOGEE (Wuk_14).Pink and purple crosses with white edges are stars from Indus and Jhelum stellar streams, respectively.Indus_13, which is an extremely -process enhanced star is tagged in the top panel.In both panels, blue (dotted), orange (dashed), and green (solid) lines are chemical evolution trajectories for galaxies with stellar masses  ★ / ⊙ of 10 6 , 10 7 , and 10 8 , respectively (Wanajo et al. 2021, see text).The solar abundance level is shown as dotted black lines in both panels.

CONCLUSIONS
We presented results from the first spectroscopic follow-up campaign for stars in the Wukong/LMS-1 dwarf-galaxy stellar stream with the Magellan Clay/MIKE combination.These targets were originally identified as members from the H3 (Naidu et al. 2020;Johnson et al. 2022) or APOGEE (Section 2.1) surveys.From these highresolution MIKE spectra, we obtained RVs, stellar parameters, and detailed chemical abundances for 14 stars in Wukong/LMS-1 cov-ering an extensive metallicity range (−3.5 < [Fe/H] ≲ −1.3).We also recalculated average abundances for a pair of stellar streams, Indus and Jhelum (Ji et al. 2020a), which have been suggested to be associated with the larger Wukong/LMS-1 (Bonaca et al. 2021;Malhan et al. 2021Malhan et al. , 2022)), in order to guarantee a consistent scale across samples.Our main results are summarized below.
• Wukong/LMS-1 is chemically indistinguishable from Indus and Jhelum.Although this is not enough to confirm that they originated in the same parent dwarf galaxy, this is certainly evidence in favor of this hypothesis.
• The [/Fe] ratios (Mg, Ca) in Wukong/LMS-1 remain high (∼0.3-0.4 dex) up to [Fe/H] ≳ −2, which is similar to relatively massive surviving dSph satellites of the Milky Way.This is in conformity with other works in the literature that estimated the mass of Wukong/LMS-1's progenitor with other methods (Malhan et al. 2021;Naidu et al. 2022).Moreover, the most metal-rich star in our sample (Wuk_14; [Fe/H] = −1.26± 0.20), as well as a Jhelum star at −1.45 dex, has lower [/Fe] by 0.1-0.2dex in comparison to the bulk of Wukong/LMS-1 stars, suggesting that this dwarf galaxy likely experienced fairly standard chemical evolution.
• Wukong/LMS-1 contains a pair of stars (Wuk_5 and Wuk_11) that are both N-and Na-rich in comparison to the bulk of the sample, which is a telltale sign that that these were born in a globular cluster with multiple stellar populations.This favors the hypothesis that Wukong/LMS-1 likely contained at least one globular cluster that has been completely disrupted.Interestingly, both of these Wukong/LMS-1 members plus a previously known N-/Na-rich Indus star all have compatible (1) metallicities ([Fe/H] ≡ 2.4).
• Because Wukong/LMS-1 is also associated with a couple of intact globular clusters, NGC 5024 (M53) and NGC 5053, plus the disrupted one, we used this information to estimate the halo virial mass of the progenitor system.For ≥3 globular clusters, a galaxy is expected to have a total mass of ≈10 10 , which corresponds to ∼1% of the present-day Milky Way.
• The [Eu/Fe] ratio in Wukong/LMS-1 stars increases as a function of metallicity within −2.5 < [Fe/H] ≲ −1.3, which can be reproduced by chemical-evolution models for similarly massive dwarfs with NSMs as the only source for the -process (Wanajo et al. 2021), i.e., without the need for a prompt source.Wukong/LMS-1 is, in this context, the first example of dwarf galaxy with -process enrichment clearly dominated by delayed sources.
This paper provides a powerful demonstration of how detailed abundances can be used to unveil the evolution of disrupted dwarf galaxies, which, given their low masses and high accretion redshift, can not be spatially resolved, or detected (depending on the exact redshift), even by JWST.The combination between Gaia and complementary spectroscopic surveys can be employed to confidently identify members of streams/substructures if one is well-informed regarding potential interlopers, in particular from GSE and Sagittarius stream.We envision that the next step for Galactic archaeology will be to obtain samples covering larger metallicity ranges for these streams/substructures.Towards the low-metallicity regime, the number of available targets drastically diminishes.In the metal-rich end, Milky Way's halo foreground contamination is difficult to deal with.Therefore, dedicated searches for these targets will be necessary for us to continue advancing our knowledge about the fundamental building blocks of our Galaxy.

Figure 2 .
Figure 2. Top panel: comparison between MIKE and survey RVs, either H3 (circles) or APOGEE (square).The vertical axis shows Δ RV , which corresponds to MIKE RVs minus H3/APOGEE ones.The black line shows Δ RV = 0 and the grey shaded region covers ±1 km s −1 , the systematic RV uncertainty for MIKE measurements.Bottom panel: log  versus  t relation (see text).Colored lines are empirical quadratic fits to various samples of low-metallicity stars(Barklem et al. 2005;Marino et al. 2008;Kirby et al. 2009;Cohen et al. 2013;Roederer et al. 2014;Jacobson et al. 2015, see Ji  et al. 2023a).The black line was used to determine the  t for the most metalpoor star analysed (Wuk_6; see text).Wuk_4 is shown with red edges in both panels (see Section 4.1).

Figure 3 .
Figure 3. Comparisons between our derived stellar parameters from MIKE spectra (horizontal axes; Table 1) and the values from the spectroscopic surveys (vertical), either H3 (circles) or APOGEE (square).Left:  eff .Middle: log .Right: [Fe/H].One-to-one lines are shown in black.Grey shaded areas represent the systematic uncertainties adopted.The red line in the right panel shows the average difference between our derived metallicities and those from H3.

Figure 4 .
Figure 4. [X/Fe] versus [Fe/H] plots for all elements estimated in this work for our Wukong/LMS-1.In this figure, Fe is always Fe i.The top right corner of each panel shows the X element represented.Filled symbols are measured abundances, while open ones are upper limits.Circles correspond to stars followed-up from H3 (Wuk_1 to _13) and the square is from APOGEE (Wuk_14).Dotted lines at [X/Fe] = 0 show the solar level.The red dashed line at [C/Fe] = +0.7 in the top left panel delineates the boundary for the definition of CEMP star (see text).The symbol with red edge is Wuk_4.In all panels, the grey dots in the background are metal-poor stars from our SAGA data base compilation (see text).

Figure 5 .
Figure 5. Abundance plots for some neutron-capture elements: [Sr, Y, Zr, Ba, La/Fe] versus [Fe/H].Yellow symbols are our Wukong/LMS-1 stars with detected abundance for the given element in each panel.Empty symbols are upper limits.The marker with red edge represents the CEMP star in our sample (Wuk_4).The yellow square is the metal-rich star from APOGEE (Wuk_14).Pink and purple crosses are stars from Indus and Jhelum stellar streams, respectively.Grey dots in the background are the metal-poor stars from the SAGA data base compilation with normal carbon enrichment ([C/Fe] ≤ +0.7).Regular CEMP stars are marked as orange diamonds ([C/Fe] > +0.7 and [Ba/Fe] ≤ +1.0).CEMP- stars ([C/Fe] > +0.7 and [Ba/Fe] > +1.0) are shown in blue triangles (see blue dashed line in the bottom left panel).The solar abundance level is shown as dotted black lines in all panels.

Figure 6 .
Figure 6.Abundance trends of -elements Mg (left and middle) and Ca (right) in [ /Fe] versus [Fe/H] format.Yellow symbols are our Wukong/LMS-1 stars.The marker with red edge represents the CEMP star in our sample (Wuk_4).The yellow square is the most metal-rich star followed-up from APOGEE (Wuk_14).Pink and purple crosses with white edges are stars from Indus and Jhelum stellar streams, respectively.In the left panel, blue (dotted), orange (dashed), and green (solid) lines are chemical evolution trajectories for galaxies with stellar masses  ★ / ⊙ of 10 6 , 10 7 , and 10 8 , respectively(Wanajo et al. 2021, see text).In the middle and right panels, the black line shows the median abundances from the SAGA data base compilation of metal-poor stars in bins of 0.25 dex in [Fe/H].The grey band represents 16 th and 84 th percentiles within the same bins.Stars from Sagittarius (Sgr), Sculptor (Scl), Fornax (Fnx), and Draco (Dra) dSph galaxies, Reticulum II (Ret II) UFD, as well as GSE disrupted dwarf, are exhibited for comparison.See text for details on our abundance compilation for these dwarf galaxies.The solar abundance level is shown as dotted black lines in all panels.The black error bars in the top right corner of the middle and right panels show statistical uncertainties alone in [Mg/Fe] and [Ca/Fe], respectively, as well as [Fe/H].
Figure 6.Abundance trends of -elements Mg (left and middle) and Ca (right) in [ /Fe] versus [Fe/H] format.Yellow symbols are our Wukong/LMS-1 stars.The marker with red edge represents the CEMP star in our sample (Wuk_4).The yellow square is the most metal-rich star followed-up from APOGEE (Wuk_14).Pink and purple crosses with white edges are stars from Indus and Jhelum stellar streams, respectively.In the left panel, blue (dotted), orange (dashed), and green (solid) lines are chemical evolution trajectories for galaxies with stellar masses  ★ / ⊙ of 10 6 , 10 7 , and 10 8 , respectively(Wanajo et al. 2021, see text).In the middle and right panels, the black line shows the median abundances from the SAGA data base compilation of metal-poor stars in bins of 0.25 dex in [Fe/H].The grey band represents 16 th and 84 th percentiles within the same bins.Stars from Sagittarius (Sgr), Sculptor (Scl), Fornax (Fnx), and Draco (Dra) dSph galaxies, Reticulum II (Ret II) UFD, as well as GSE disrupted dwarf, are exhibited for comparison.See text for details on our abundance compilation for these dwarf galaxies.The solar abundance level is shown as dotted black lines in all panels.The black error bars in the top right corner of the middle and right panels show statistical uncertainties alone in [Mg/Fe] and [Ca/Fe], respectively, as well as [Fe/H].

Figure 7 .
Figure 7. Top: [Na/Fe] NLTE versus [N/Fe].Bottom: [Na/Fe] NLTE versus [Fe/H].In both panels, [Na/Fe] NLTE represents our sodium-to-iron ratios corrected with the NLTE departures from Lind et al. (2011, see text).Yellow symbols are Wukong/LMS-1 stars in our sample.The marker with red edge represents the CEMP star in our sample (Wuk_4).The yellow square is the most metal-rich star followed-up from APOGEE (Wuk_14).Star symbols represent the N-/Na-rich stars (Wuk_5and Wuk_11).Transparent symbols without edges in the background correspond the same Wukong/LMS-1 stars, but without the NLTE corrections to Na. Pink and purple crosses with white edges are stars from Indus and Jhelum stellar streams, respectively.Indus_0, which is also N-/Na-rich, is tagged in both panels.The solar abundance level is shown as dotted black lines in both panels.

Figure 8 .
Figure 8. Top: [Eu/Fe] versus [Fe/H].Bottom: [Eu/Mg] versus [Mg/H].Yellow symbols are Wukong/LMS-1 stars in our sample.Empty symbols are upper limits.The marker with red edge represents the CEMP star in our sample (Wuk_4).The yellow square is the most metal-rich star followed-up from APOGEE (Wuk_14).Pink and purple crosses with white edges are stars from Indus and Jhelum stellar streams, respectively.Indus_13, which is an extremely -process enhanced star is tagged in the top panel.In both panels, blue (dotted), orange (dashed), and green (solid) lines are chemical evolution trajectories for galaxies with stellar masses  ★ / ⊙ of 10 6 , 10 7 , and 10 8 , respectively(Wanajo et al. 2021, see text).The solar abundance level is shown as dotted black lines in both panels.

Table 1 .
Observational information, RVs, and stellar parameters for Wukong/LMS-1 stars analysed in this work.The reported / values are per pixel.Model atmosphere metallicity values ([M/H]) can be larger than [Fe/H] by up to 0.03 dex.