The heavy-elements heritage of the falling sky

Recent dynamical analysis based on Gaia data have revealed major accretion events in Milky Way's history. Nevertheless, our understanding of the primordial Galaxy is hindered because the bona fide identification of the most metal-poor and correspondently oldest accreted stars remains challenging. Contrary to alpha-elements, neutron-capture elements present unexplained large abundance spreads for low metallicity stars, that could result from a mixture of formation sites. We have analysed the abundances of yttrium, europium, magnesium and iron in Milky Way satellite galaxies, field halo stars and globular clusters. The chemical information has been complemented with orbital parameters based on Gaia data. In particular, the orbit's average inclination has been considered. The [Y/Eu] abundance behaviour with respect to the [Mg/Fe] turnovers for satellite galaxies of different masses reveals that higher luminosity systems, for which the [Mg/Fe] abundance declines at higher metallicities, present enhanced [Y/Eu] abundances, particularly in the [Fe/H] regime between -2.25 and -1.25 dex. In addition, the analysis has uncovered a chemo-dynamical correlation for both globular clusters and field stars of the Galactic halo, accounting for about half of the [Y/Eu] abundance spread. [Y/Eu] under-abundances typical of protracted chemical evolutions, are preferentially observed in polar-like orbits, pointing to a possible anisotropy in the accretion processes. Our results strongly suggest that the observed [Y/Eu] abundance spread in the Milky Way halo could result from a mixture of systems with different masses. They also highlight that both nature and nurture are relevant to the Milky Way formation, since its primordial epochs, opening new pathways for chemical diagnostics of our Galaxy building up.


Introduction
The most primitive Galactic stars are essential to understand the Milky Way formation. Nevertheless, the robust identification of accreted objects is particularly challenging for stars with primordial abundances having at most 30 times less metals than the Sun ([Fe/H] -1.5). Kinematical or dynamical indications of accretion are insufficient to reveal ancient mergers (Jean-Baptiste et al. 2017). They need to be complemented by chemical diagnostics (Freeman & Bland-Hawthorn 2002), as the chemical evolution of a system strongly depends on its mass. Compared to the massive Milky Way, satellite galaxies generally present signs of protracted evolutions, being more metal deficient and showing a variety of chemical patterns that we should retrieve in the accreted populations, now mixed with in situ formed stars.
The most commonly used chemical diagnostic of accretion is the α-elements (O, Mg, Si, S, Ca, Ti) ratio with respect to iron ([α/Fe]). Initially enhanced, the [α/Fe] abundance starts to strongly decline with metallicity after the supernovae Ia explosion rate reaches a maximum (Matteucci & Greggio 1986). This produces a knee in the [α/Fe] vs. [Fe/H] trend whose location provides constraints on the system total mass: the less massive the system, the more metal-poor is the [α/Fe] turnover. Unfor-tunately, this accretion diagnostic is not discriminating enough for stars in the Galactic halo, with metallicities lower than the [α/Fe] turnover of most satellite galaxies. As a consequence, metal-poor field stars kinematically proposed to be members of ancient accreted satellites, like Gaia-Enceladus/Sausage (Helmi et al. 2018;Belokurov et al. 2018), have similar [α/Fe] abundances as non-members for [Fe/H] -1.5 dex. They only appear as a separate sequence at higher metallicity (Helmi et al. 2018), hampering also the detection of low mass mergers. Similarly, the population of clusters in the Galactic halo is mostly homogeneous in their [α/Fe] abundances (Recio-Blanco 2018).
Galactic Archaeology thus needs a new accretion diagnostic to understand the primordial stellar populations and, in this work, we have used neutron-capture elements to identify it. Contrary to α-elements, neutron-capture elements present unexplained large abundance spreads for low metallicity stars, that could result from a mixture of formation sites. In particular, we have considered the logarithm of the ratio of a star's yttrium abundance with respect to its europium one, [Y/Eu]. Approximately 75% of the solar Yttrium was produced (Prantzos et al. 2018) by low and intermediate mass asymptotic giant branch (AGB) stars, through slow neutron captures (relatively to the βdecay rates of unstable nuclei). In addition, first peak s-elements Article number, page 1 of 6 arXiv:2007.08313v1 [astro-ph.GA] 16 Jul 2020 A&A proofs: manuscript no. RecioBlanco like Y have a larger contribution from low mass stars than second peak elements like Ba. On the other hand, 94% of europium is produced by massive stars through rapid neutron captures (Bisterzo et al. 2014). Proposed Eu production sites are neutron star mergers (Rosswog et al. 1999), high energy winds accompanying core collapse supernovae explosions (Woosley et al. 1994) or magneto-hydrodynamical explosions of fast rotating stars (Winteler et al. 2012). As a consequence, the [Y/Eu] abundance ratio characterizes the relative contribution of low-intermediate mass stars with respect to high mass stars, being therefore a good indicator of the chemical evolution efficiency.

Chemical abundances and orbital parameter estimations
The present study relies on several samples of objects: globular clusters and field stars, both from the Milky Way and its satellites. They have been computed as the average values over 10 Gyr of integration. To this purpose, we used the median values obtained from 1000 orbits for each cluster obtained through Monte Carlo realizations of the initial conditions, considering the observational measurements and their errors. In particular, the orbit's average inclination has been computed as arccos(Lz/Ltot). In our convention, the orbital inclination is defined from the Galactic plane and comprised between 0 • and 180 • , with prograde orbits below 90 • . Error bars in the orbital parameters associated to model assumptions, have been estimated by comparing the results obtained with different Galactic potentials (defined as Model-1, -2, -3 in Gaia Collaboration et al. 2018b). In particular, the dispersion in the orbital inclination (estimated as the third quantile value of the differences distribution between two models) is 6 degrees. In addition to this main dataset of cluster orbits, we have completed the sample with six additional objects from Vasiliev (2019).
For our field stars samples, we have derived the orbital parameters using the python package galpy (Bovy 2015). We assume the MWPotential14 Milky Way mass model included in this package. We derived the action parameters through the action-angle isochrone approximation (Bovy 2014). As input parameters we have used the radial velocities gathered in Simbad, the Gaia DR2 proper motions and the distances from Bailer-Jones et al. (2018). In addition, we have checked the effect of using two different methodologies of the dynamical parameters for clusters and field stars. To this purpose, we have re-computed the clusters orbital inclinations using the field stars methodology calculated the differences with respect to the Model-2 orbital results from Gaia Collaboration et al. 2018. The median absolute deviation of the orbital inclination differences is 2.5 degrees, confirming the consistency of the two approaches.
Finally, we have assessed the impact of the detected Gaia kinematic biases (Schönrich et al. 2019)     Panels c and d show the normalized distribution of orbital inclinations for the three sets of objects, selected either with the [Y/Eu] diagnostic or with the [Mg/Fe] one, respectively. Although the two chemical diagnostics target different objects (those in common being excluded from panel d histograms) and span different metallicity regimes, the similarities between panels c and d distributions are important. Twosampled Kolmogorov-Smirnov tests between the nine possible pairs of distributions have been performed to test this similarity. The null hypothesis, assuming that the samples come from a population with the same distribution, is rejected for all the pairs except those having the same colour (targeting therefore the same parent system mass abundances (redgreen and green-blue pairs), partially overlap in their orbital inclination distributions as a result of abundance uncertainties, but also to the fact that no perfectly separated components seem to exist. In particular, in situ formed objects dynamically heated by past mergers (e.g. Belokurov et al. 2019;Di Matteo et al. 2019) could also blur the orbital inclination distributions.

Chemo-dynamical correlations and abundance spread in the Halo
The above result confirms the coherence of the [Y/Eu] diagnostic with the [Mg/Fe] one, revealing possible chemodynamical correlations with two independent chemical indicators. To quantify those trends, Figure 3 shows the deviations in [Y/Eu] and [Mg/Fe] abundances with respect to the average, as a function of orbital inclination. Contrary to the analysis of Figure 2, no data subsamples are predefined and the considered metallicity regime spans -2.0 ≤ [M/H] ≤-1.2 dex in both panels. The two chemical diagnostics show under-abundances around the polar direction (60 • inclination 120 • ) and overabundances near the plane (prograde objects with inclination 60 • and retrograde objects with inclination 120 • ). The observed chemo-dynamical correlations, including both globular clusters and field stars, are more pronounced for the [Y/Eu] abundances than for the [Mg/Fe] ones as expected from their corresponding abundance spreads in this metallicity regime. In particular, the orbital inclination seems to account for about half of the [Y/Eu] abundance scatter.

Conclusions
Although Galactic studies need to be constantly validated in the huge parameter space of Milky Way populations, the observed chemo-dynamical correlations open new paths of exploration of our Galaxy formation history. In the light of the previous conclusions, the heavy elements abundance scatter of the primordial Milky Way possibly results from an amalgam of systems with different masses and chemical evolutions. First, objects in polar-like orbits showing underabundances of [Y/Eu] could result from a composite debris from low mass accretions. Interestingly, polar orbits are also found for more recent merger events as the Sagittarius one. This suggests the possible existence of a preferential accretion axis around the polar direction, linking the Milky Way to its satellites and deserving further study. In the metal-poor and intermediate metallicity regime, where the [Y/Eu] under-abundances are larger than the [α/Fe] ones, future large scale heavy-element studies seem crucial to distinguish between low-mass accretions and slow rotating debris from more massive mergers.
Second, satellite merger debris in retrograde orbits was previously suggested by the analysis of several dynamical overdensities (e.g. Helmi et al. 2018;Belokurov et al. 2018;Myeong et al. 2019), and attributed to high mass progenitors (Gaia Enceladus/Saussage, Sequoia). In our study, the chemical patterns dominating that retrograde regime near the plane are indeed typical of high mass systems, reaching metallicities of -0.5 dex and relatively high [Y/Eu] abundances. The interplay of this old retrograde population with the prograde disc and the slow rotating accretion debris is probably an important piece of the Galaxy formation puzzle.
Third, a prograde population, showing [Y/Eu] overabundances, seems to be present even in the low metallicity regime. It could be the fossil signature of the primitive collapsed Galaxy, probably occupying prograde orbits near the plane, as the more metal-rich disc. This hypothesis is strengthen by the recent discovery of very metal-poor stars with disc like orbits (Sestito et al. 2020) In conclusion, both nature and nurture appear to have played a role to build up the ancient Milky Way, leaving inprints we are starting to decode. Chemical diagnostics, including heavy elements abundances, will certainly be fundamental in the on going Gaia revolution.
Acknowledgements. This work has made use of data from the European Space Agency (ESA) mission Gaia Data Processing and Analysis Consortium (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. ARB, PdL and EFA acknowledge financial support from the ANR 14-CE33-014-01. TA has received funding from the European Union's Horizon 2020 research and innovation programme under Marie Sklodowska-Curie grant agreement number 745617 and also acknowledges funding from the MINECO (Spanish Ministry of Economy) through grants ESP2016-80079-C2-1-R (MINECO/FEDER, UE) and ESP2014-55996-C2-1-R (MINECO/FEDER, UE). AH acknowledges funding from a Vici grant from the Netherlands Organisation for Scientific Research (NWO). We thank E. Vasiliev for providing his orbital parameters for globular clusters. ARB thanks Vanessa Hill, Sebastian Peirani and Oliver Hahn for useful discussions and Chris Wegg for kindly language corrections.