High speed stars: III. Detailed abundances and binary nature of the extreme speed star GHS143 ⋆

Context. The Gaia satellite has provided the community with three releases containing astrometrical and photometric data as well as by products, such as stellar parameters and variability indicators. Aims. By selecting in the Gaia database, one can select stars with the requested characteristics, such as high speed. At present any selection is based on available Gaia releases including a subset of the observations. This, for some stars, can show some limitations, for example there is still not a su ffi cient number of observations to detect binarity. Methods. We investigated a star selected in Gaia EDR3 for its high speed that appears unbound to the Galaxy. We requested high-quality spectra to derive more information on the star. Results. From the spectroscopic investigation we confirm the low metallicity content of the star, and we derive a detailed chemical composition. The star is poor in carbon and very rich in oxygen: [(C + N + O) / Fe] =+ 0.65. From the two spectra observed we conclude that the star is in a binary system and from the investigation of the ionisation balance we derive that the star is closer than implied by the Gaia DR3 parallax, and thus has a a lower intrinsic luminosity. Conclusions. The star is probably still unbound, but there is the possibility that it is bound to the Galaxy. Its low carbon abundance suggests that the star was formed in a dwarf galaxy.


Introduction 1
In an ongoing investigation, we are trying to characterise the 2 population of stars with high speed with respect to the Sun.In Based on observations made with UVES at VLT 111.24J1.001and 112.25EH.001.
The lines investigated are on an on-line table at CDS. evolved star of G magnitude 13.06 and of apparent young age, characterised by extreme kinematics.Assuming the parallax, proper motions, and radial velocity in Gaia DR3 (Gaia Collaboration et al. 2022), it can be seen that GHS143 is not bound to the Galaxy, but is falling into it.We requested high-resolution spectra of this star to investigate if the stellar parameters we derived from a high-resolution spectrum are consistent with the Gaia DR3 parallax and photometry, and also to check for possible radial velocity variations.In this paper we use these spectra to derive a complete chemical inventory of this star, derive the uncertainties, and discuss the possibility that the star is bound to the Galaxy.

Observations
Two UVES (Dekker et al. 2000) spectra have been secured for this star.In the ESO programme 111.22EH.001 the star was observed on August 20, 2023, in the setting DIC2 437+760 (wave-A&A proofs: manuscript no.output

Stellar parameters
The star was analysed in Paper II and the stellar parameters adopted were T eff =5159 K, log g=1.8 dex, a microturbulence ξ = 1.96 kms −1 ; and an iron abundance of [Fe/H] = −1.74dex was then derived.When adopting these stellar parameters to analyse the UVES spectra, we obtain Fe abundances in very good agreement ([Fe/H] = −1.86 ± 0.10 dex and [Fe/H] = −1.85± 0.10 dex from the two spectra), and in both cases a good Fe ionisation balance.
We allowed MyGIsFOS to derive the parameters and we obtained for T eff , log g, ξ, and [Fe/H]: 4988 ± 96, 1.37 ± 0.05, 1.59 ± 0.07, and −1.92 ± 0.10 from the spectrum of August and 4956 ± 93, 1.40 ± 0.07, 1.54 ± 0.08, and −1.95 ± 0.12 from the spectrum of November.The two spectra provide very coherent results.We used the calibration suggested by Frebel et al. (2013) to bring the effective temperature derived by the excitation on the photometric scale.With this calibration we derived 5159 and 5127 K, respectively, in excellent agreement with the value derived from the Gaia DR3 photometry and applied in Paper II.By using the calibration by Mucciarelli & Bonifacio (2020) to derive T eff on the photometric scale of González Hernández & Bonifacio (2009), we derived 5136 and 5112 K, respectively, with an uncertainty of 130 K.We adopted T eff = 5160 K.
To derive the stellar parameters we focused on the spectrum from November 2023, which has a higher flux, and to derive the uncertainties we used the August 2023 spectrum.We are aware that the Fe abundance derived from Fe i lines is affected by departure from local termodymanycal equilibrium (hereafter NLTE, the local termodymanycal equilibrium shall be referred to as LTE) and forcing the ionisation equilibrium does not take into account the NLTE effects.The NLTE correction we expect for this star is about 0.1 dex.We selected the Fe i lines retained by MyGIsFOS and also available from the web site of MPIA, 1 and with these 39 Fe i lines (providing [Fe/H] = −1.76)we derived the NLTE corrections (Bergemann et al. 2012b).For a sample of 30 Fe i lines with NLTE corrections from (Bergemann et al. 2012b), we verified the NLTE corrections as 3D NLTE − 1D LTE in Amarsi et al. (2016) and, with exactly the same stellar parameters, the average difference is 0.02 dex.We gave as input to MyGIsFOS several log g values (see Fig. 2) with fixed T eff ; we derived the best agreement in [Fe/H] from the Fe i and Fe ii lines for log g = 2.1 dex.The NLTE correction is of 0.11 dex for log g = 2.1 and spans values from 0.08 dex for the highest surface gravity to 0.13 dex for the lowest.These corrections are applied in Fig. 2. With fixed effective temperature and assuming LTE, the iron ionisation balance implies log g = 1.8, as expected, lower than the value implied by the NLTE iron abundances.
We fixed T eff and log g to derive the microturbulence and derived 1.65 ± 0.07 kms −1 , changing very little by changing log g (a change of 0.02 kms −1 for a variation of 0.4 dex in log g).The microturbulence derived by using the calibration of Mashonkina et al. ( 2017), with T eff =5159 K and log g=2.1 dex, provided 1.83 kms −1 .By assuming T eff =5159 K, log g=2.1 dex, and ξ = 1.65 kms −1 , as derived by MyGIsFOS, we derived [Fe/H] of −1.81 ± 0.11 and −1.70 ± 0.11, when derived from Fe i and Fe ii lines, respectively, in perfect agreement if we expect an NLTE correction of 0.11 dex (Bergemann et al. 2012b) on [Fe/H] from Fe neutral lines.In Fig. 2 we show how the iron abundance derived from Fe i and Fe ii lines changes, as a function of the adopted log g.All values of surface gravity in the range 1.8-2.5 are acceptable, considering the involved uncertainties.

Stellar mass
In Paper II, we adopted the parallax provided in Gaia DR3 corrected by the zero-point (Lindegren et al. 2021), and we derived a mass of 3.1 M or 3.8 M .According to the Fe ionisation balance, from the UVES spectra, we favour a higher log g that would imply that the star has a lower intrinsic luminosity, so it has to be closer with a larger parallax.In spite of the uncertainties that plague the spectroscopic surface gravity determination, we believe that in this case it is more reliable than that derived from the parallax.Since our radial velocity measurements imply that the star is a binary, we expect that its astrometric measurements also contain a component due to the orbital motion.The astrometric data should then be processed using one of the astrometric binary processing pipelines of Gaia (Halbwachs et al. 2023;Holl et al. 2023) that would result in a parallax different from that available in Gaia DR3, which was obtained treating the star as a single star.As mentioned above, the spectroscopic surface gravity implies that this parallax should be larger than that in Gaia DR3.With the adopted parameters (T eff =5159, log g=2.1, [Fe/H]=-1.8),we derived these possible masses and ages (see In the most extreme case, with an adopted log g of 2.5 dex, we derived the following: -M = 1.4M and age of 2 Ga (3); -M = 0.8M and age of 8.4 Ga (7).

Kinematics
The fact that the star is a spectroscopic binary implies that the Gaia DR3 parallax may be incorrect.The Gaia DR3 parallax was derived assuming that GHS143 is a single star.The quality of the UVES spectra allowed us to put strong constraints on the Fig. 3. GHS143 (red circle) in the colour magnitude diagram compared to two isochrones at metallicity −1.65 highlighting the ranges for possible solutions: 500 Ma (black and pink, the evolution stages for the first two solutions) and 1 Ga (dark blue and light blue, the evolution stage for the third solution).The solution with the Gaia DR3 parallax (large green square) adopted in Paper II is compared to the isochrone with an age of 200 Ma (green points).
surface gravity of the star, based on the iron ionisation equi-171 librium, for which we took into account the NLTE effects.Our 172 preferred gravity is log g = 2.1; taking into account the errors 173 on the Fe i and Fe ii abundances, any surface gravity in the in-174 terval 1.8 -2.5 is consistent with the observations.If we turn 175 around the Stefan-Boltzmann equation and associate a paral-176 lax with each surface gravity, this translates into parallaxes from 177 0.098 mas to 0.222 mas.We investigated how the dynamics of 178 the star changes for various parallaxes in this range.For each 179 parallax we proceeded as in Paper II; we employed the galpy 180 code and the MWPotential2014 Galactic potential (Bovy 2015) 181 and the same assumptions on the solar position and motion.For 182 each parallax we considered the astrometric covariance matrix 183 and used the Pyia code (Price-Whelan 2018) to produce a ran-184 dom realisation of the stellar kinematic data.For each parallax 185 we extracted 1000 realisations and used them as input to galpy 186 to evaluate the dynamical status of the star.

187
For all parallaxes smaller than 0.168 mas, corresponding to 188 log g = 2.24, the star is unbound, as derived by Bonifacio et al. 189 (2024) from the Gaia DR3 parallax.For a parallax of 0.168 mas 190 the star is partially unbound, in the sense that it is unbound for 191 609 realisations out of 1000.For larger parallaxes the star be-192 comes bound, albeit with a large apocentre, in excess of 30 kpc 193 with our adopted Galactic potential.We conclude that the bound-194 ary between being bound and unbound is around log g = 2.2, 195 higher surface gravities make the star bound, while lower values 196 make it unbound, in the adopted Galactic potential.We did not detect any 13 C (see Fig. 4), and we concluded that 203 the 12 C/ 13 C is not higher than solar.We investigated the CN The Cu i line at 578.2 nm provides a negative [Cu/Fe] ratio ([Cu/Fe] = −0.37 dex), but this is known to be a NLTE effect (see Caffau et al. 2023).The star is slightly enhanced in Zn ([Zn/Fe] = 0.19 dex) and in Zr ([Zr/Fe] = 0.34 dex).For Zn, according to Sitnova et al. (2022), the NLTE correction on A(Zn) is 0.16 dex, so taking into account NLTE corrections on both elements, we derive [Zn/Fe] = 0.24.
We investigated two Sr ii (407.7 and 421.5 nm) lines and one Sr i (460.7 nm) to derive the Sr abundance.The NLTE correction of the Sr i is about 0.4 dex according to Bergemann et al. (2012a).We fitted the five Ba ii lines (455.4, 493.4, 585.3, 614.1, and 649,6 nm) available in the wavelength range and we derived that the star is slightly enhanced in Ba ([Ba/Fe] = 0.43) (see Fig. A.3).We verified that the partition function for La, Ce, Nd, Sm, Eu, Dy, and Er in SYNTHE was good by computing few lines with Turbospectrum and derived very consistent results.For the heavy elements we investigated the La ii, Ce ii, Nd ii, Sm ii, Dy ii, and Er ii lines.The two Eu ii lines (412.9 and 664.5 nm) were fit and they provided [Eu/Fe]=0.25.

Uncertainties
The uncertainties in the stellar parameters are related to the uncertainties in the Gaia photometry and astrometry, the way to derive the stellar parameters, and the reddening.In Sect.3.2, different ways to derive T eff are discussed, each of which brings very consistent values.Had we used the calibration by Mucciarelli et al. (2021) instead of the comparison to synthetic photometry and colour as applied in Paper II, we would have derived a temperature 60 K hotter.We then assume an uncertainty of 100 K in T eff .The microturbulence derived from the Fe i lines provides a value of 1.65 kms −1 , while the calibration by Mashonkina et al. (2017) provides a value about 0.2 kms −1 higher.We assigned 0.2 kms −1 as the uncertainty in microturbulence.
For the surface gravity things are more complicated.The uncertainty in the parallax is of 37%, providing an uncertainty in log g of 0.2 dex.However, deriving the log g from the balance of A(Fe) from the Fe i and Fe ii lines, we converge to a log g of 2.1 dex, which is 0.3 dex higher than the value derived from the parallax.With log g=1.8 dex as derived from the Gaia DR3 photometry and parallax corrected by the zero point, we obtained a perfect balance of A(Fe) from the Fe i and Fe ii lines in LTE, and the value would still be compatible after applying the NLTE correction within the uncertainties.With log g=2.5 dex we derived A(Fe) = −1.81± 0.11 from the Fe i lines to which we have to add 0.08 dex to take into account NLTE corrections and A(Fe) = −1.52 ± 0.12 from the Fe ii lines.The two values, −1.73 ± 0.11 and −1.52 ± 0.12 dex, are compatible within the uncertainties.We adopted 0.4 dex as the uncertainty in the surface gravity (see Fig. 2).
The uncertainties in the stellar parameters implies uncertainties in the abundances derived, and are provided in Table A.2.To derive the random uncertainties we compared the results of the two UVES spectra.

Discussion and conclusions
The analysis of the UVES high-resolution spectra of GHS143 has allowed us to gain further insight into the nature of this star, and also to highlight some difficulties in the interpretation of the data.The first important result is that the star is a single-spectrum spectroscopic binary (SB1), on the basis of the radial velocities measured from the two spectra.As discussed in Sect.3.4, the iron ionisation equilibrium allows a range of possible surface gravities.The range 2.2 ≤ log g ≤ 2.5 implies a distance that makes the star bound.The young apparent age and high mass of the star remain true for all but the extreme gravity of log g = 2.5 dex and the assumption that it is on the AGB.Even in this extreme case, however, the age is only 8.4 Ga, which is younger than the bulk of the halo stars.This, coupled with the large apogalacticon distances implied in all cases that make the star bound, makes it more likely that the star formed in an external galaxy and was accreted by the Milky Way.We propose that GHS143, whether bound or unbound, was formed in an external galaxy.We think that the peculiar CNO abundance pattern of GHS143 is a specific signature of this galaxy, although we cannot point to any example of a galaxy with such a chemical pattern.
Acknowledgements.The authors wish to thank the referee.We gratefully acknowledge support from the French National Research Agency (ANR) funded projects "Pristine" (ANR-18-CE31-0017).This work has made use of data from the European Space Agency (ESA) mission Gaia (https://www.cosmos.esa.int/gaia),processed by the Gaia Data Processing and Analysis Consortium (DPAC, https://www.cosmos.esa.int/web/gaia/dpac/consortium).Funding for the DPAC has been provided by national institutions, in particular the institutions participating in the Gaia Multilateral Agreement.This research has made use of the SIMBAD database, operated at CDS, Strasbourg, France.

Fig. 1 .
Fig. 1.Observed spectra in the Hα region.The spectrum with higher flux is from November 2023; the lower flux is from August 2023.

Fig. 2 .
Fig. 2. [Fe/H] derived from Fe i (filled black squares) and corrected by NLTE and from Fe ii (blue open circles) lines vs log g.

Fig. 3
Fig. 3): -M = 2.3M and age of 493 Ma (3, RGB, red giant branch, or the quick stage of red giant for intermediate+massive stars 2 ); -M = 2.3M and age of 504 Ma (4, CHEB, core He-burning for low mass stars, or the very initial stage of CHeB for in-termediate+massive stars); -M = 1.9M and age of 904 Ma (7, EAGB, the early asymptotic giant branch).

198
The chemical investigation provided inTable A.1 is from the 199 spectrum observed in November 2023.The star is poor in car-200 bon, but rich in nitrogen and oxygen.The C abundance was 201 derived by line profile fitting of the G-band at about 428 nm.202

Fig. 4 .
Fig. 4. Observed spectra (solid black) compared to theoretical synthesis in the G-band range where 13 C line are expected.
204 band at 383 nm and, fixing the C abundance, we derived N abun-205 dance of A(N)=6.83.Oxygen was derived from the [OI] line at 206 630 nm and from the triplet at 777 nm.The lines of the triplet 207 are affected by non-negligible NLTE effects, while the forbid-208 den line forms in conditions close to LTE.By correcting the O 209 abundance for NLTE effects (Sitnova et al. 2013) the four O i 210 lines are in very close agreement, and in this case we derived 211 [(C+N+O)/Fe]=0.65.212 The star is enhanced in α elements.From ten Mg i lines we 213 derive [Mg/Fe] = +0.52 (see Fig. A.3).The NLTE corrections 214 for Fe and Mg are comparable (+0.08 dex for the four Mg i lines 215 used here; Bergemann et al. 2017), so the [Mg/Fe] ratio, taking 216 into account the NLTE corrections, is close to the LTE value.217 From 15 lines of Si i we derived [Si/Fe] = +0.46 and from one respectively) provide consistent values with Fe (see Fig. A.3).
The abundances derived for this star seem well in line with the typical abundance patterns found in the Milky Way halo for all elements except CNO.The pattern of the CNO abundances is quite exceptional.A carbon-to-iron ratio [C/Fe] ≈ −0.4 is compatible with what is observed in some ultra-faint dwarf galaxies such as Boo I (see e.g.Frebel et al. 2016;Norris et al. 2010, and references therein), Segue 1 (Norris et al. 2010), Uma II, and Coma Ber (Frebel et al. 2010), and also in the Milky Way halo (see e.g.Barklem et al. 2005).However, a carbon-to-oxygen ratio [C/O]=-1.2 is far lower than what is observed in Milky Way stars (Akerman et al. 2004; Spite et al. 2005).There are unfortunately not enough oxygen measurements in dwarf galaxies to say much about the C/O ratio.The ratio [(C+N+O)/Fe]=+0.65 is very high and quite exceptional.While it may be tempting to interpret the low [C/O] ratio, coupled with the high [N/O] = −0.03,as the result of CNO processing, this is not possible in view of the robustly established high 12 C/ 13 C ratio.The N/O ratio observed in this star is compatible with the ratio observed in Galactic H ii regions at galactocentric distances of 10-14 Kpc (Arellano-Córdova et al. 2021), although both the nitrogen and oxygen abundances are almost 1 dex lower.

Fig
Fig. A.1.O i triplet in the observed spectrum of November 2023 (solid blck) compared to a synthesys (solid red) with A(O) value average from the O i triplet lines.