The Milky Way appears to have formed in three kinematic-chemical phases — a chaotic protogalaxy, a kinematically hot high-α 'old' disk, and a kinematically cold low-α 'young' disk — with the Gaia-Sausage-Enceladus (GSE) major merger, roughly 10 Gyr ago, proposed as the decisive event linking them. Chandra et al. (2024), using 10 million Gaia red giants and an analog from the TNG50 cosmological magnetohydrodynamical simulation, concluded that this merger-driven picture is supported by the data but not yet demanded by it. The open problem is to establish or refute the causal, merger-driven interpretation by combining precise stellar ages, chemical abundance patterns, and improved simulations. Its linked fronts are: (i) the temporal ordering of the truncation of the high-α disk, the arrival of the GSE merger, and the birth of the low-α disk; (ii) whether the [α/Fe]–[Fe/H] bimodality (the split of disk stars into two abundance sequences) is generic to external Milky-Way-like galaxies and whether a major merger is required to produce it; (iii) the relative importance of intrinsic birth conditions versus post-formation heating in setting the thick disk's vertical structure; and (iv) whether the in-situ eccentric 'Eos' stellar population arises naturally in hydrodynamic simulations of the MW–GSE encounter. Recent work contests the merger interpretation itself (a bar-driven origin of the bimodality; a possible absence of the bimodality in M31).

Background

The European Space Agency's Gaia mission has delivered astrometry and low-resolution (XP) spectrophotometry for over a billion stars, which for roughly 10 million red giant stars can be turned into metallicities [Fe/H] (iron abundance relative to solar, in logarithmic units) and α-abundances [α/Fe] (abundance of α-process elements such as magnesium and silicon, synthesized in core-collapse supernovae, relative to iron, which is produced mainly by type Ia supernovae on longer timescales). When Milky Way disk stars are plotted in the [α/Fe]–[Fe/H] plane they split into two sequences: a metal-poor, α-enhanced population associated with the kinematically hot, geometrically thick disk, and a metal-rich, low-α population associated with the kinematically cold, geometrically thin disk. This 'chemical bimodality' is a central datum of Galactic archaeology, and the two sequences are commonly interpreted as an early, intense star-formation episode followed by a more quiescent one.

Kinematically, Chandra, Semenov, Rix, Conroy and colleagues (2024) analyzed the angular momentum of these Gaia giants as a function of metallicity and identified three phases of Galactic evolution: a disordered protogalaxy, an 'old' high-α disk that spins up from the protogalaxy and then gradually cools down toward more circular orbits, and a 'young' low-α disk that is born kinematically cold. Orbital circularity (the ratio of a star's angular momentum to that of a circular orbit of the same energy) serves as the key kinematic diagnostic. In their picture the transition between the two disks is driven by the last major merger of the Milky Way: the Gaia-Sausage-Enceladus (GSE) event, identified from the Gaia era discovery that most of the Galaxy's stellar halo consists of stars on eccentric, low-angular-momentum orbits deposited by a single satellite with roughly 10% of the Milky Way's mass. Star formation in that satellite is inferred to have truncated 9.5–10 Gyr ago (redshift z ≈ 2), which dates the merger. The merger would have heated and tilted the ancient disk (creating the 'splash' or in-situ halo, stars on halo orbits whose chemistry matches the old disk), while delivering pristine, low-metallicity gas from which the low-α thin disk subsequently formed — a 'metallicity reset' that naturally produces the observed bimodality.

Chandra et al. supported this scenario with a Milky Way analog (dark-matter halo #519311) from the TNG50 cosmological magnetohydrodynamical simulation, the only one of the 61 Milky-Way-like TNG50 halos reproducing the observed three-phase behavior in the circularity–metallicity plane. Notably, this analog's gas-rich major merger occurred only ~7–8 Gyr ago, slightly later than the inferred GSE epoch — a discrepancy the source authors note as a caveat rather than as one of their open questions — and the outcome of mergers is known to depend on simulation subgrid feedback prescriptions (numerical recipes for stellar feedback acting below the simulation's resolution scale; e.g., simulations from the FIRE project have burstier star formation than TNG). A further motivation is the 'Eos' population found by Myeong et al. (2022) in an unsupervised chemodynamical decomposition (data-driven clustering of stars in the combined space of chemistry and kinematics) of the local stellar halo: an in-situ (born inside the Milky Way), low-α, metal-poor component on eccentric orbits that chemically connects the GSE debris with the low-α disk, plausibly the earliest thin-disk stars born from disk gas polluted by GSE — but this identification is untested.

The picture's weak points, as stated by the source authors themselves, are: precise stellar ages for large samples are still lacking, so the temporal ordering of high-α disk truncation, GSE arrival, and low-α disk birth remains disputed (different studies, e.g., Bonaca et al. 2020 and the age-dating survey of Xiang & Rix 2022, arrive at different orderings); it is unknown how common the [α/Fe]–[Fe/H] bimodality is in external Milky-Way-like galaxies (integral-field spectroscopic surveys — which record a spectrum simultaneously at many spatial positions across a galaxy — such as GECKOS are targeting this, and first measurements with the James Webb Space Telescope (JWST) in M31 exist); the relative importance of thick birth orbits versus later dynamical heating for the thick disk's vertical structure remains debated and feedback-implementation dependent; and whether Eos arises in hydrodynamic simulations of the MW–GSE encounter has not been tested. Separately, the authors note that the analog's gas-rich merger is 'slightly later' than the hypothesized timing of the Milky Way's own GSE merger, a caveat they note but do not raise among their open questions. Hence the formation picture is 'supported — but not demanded — by current data'.

Problem Statement

Can the Milky Way's three-phase formation picture — a chaotic protogalaxy, spinup into a kinematically hot high-α old disk, and the Gaia-Sausage-Enceladus (GSE) major merger ~10 Gyr ago delivering the gas from which the kinematically cold low-α young disk formed — be made definitive, i.e., established or refuted as demanded by the data rather than merely supported by one simulation analog, as Chandra et al. (2024) framed it? Answering this overarching question requires, as jointly supporting and scientifically linked deliverables: (i) a robust temporal ordering, from precise stellar ages, of the truncation of the high-α disk, the arrival of the GSE merger, and the birth of the low-α disk, so that the causal links between the merger and the disk transitions can be tested rather than assumed; (ii) a determination of whether the [α/Fe]–[Fe/H] chemical bimodality is generic to external Milky-Way-like galaxies, that is, whether it is present across a sample of such galaxies (of which M31 is the first tested case), and whether a GSE-like major merger is required to produce it; (iii) a determination of the relative importance of intrinsic birth conditions versus post-formation dynamical heating in setting the vertical structure of the thick disk, including how the answer depends on simulation feedback prescriptions; and (iv) a test of whether an in-situ, eccentric, low-α population like Eos arises naturally in hydrodynamic simulations of the MW–GSE encounter, and whether its predicted kinematics match the observed Eos stars. A complete answer demonstrates, or rules out, the causal merger-driven interpretation of the three-phase evolution by combining precise ages, detailed chemical abundance patterns, and improved simulation predictions; partial answers that settle individual fronts are valuable but do not by themselves discharge the question.

References

  1. Chandra, V., Semenov, V. A., Rix, H.-W., Conroy, C., Bonaca, A., Naidu, R. P., Andrae, R., Li, J., & Hernquist, L. 2024, 'The Three-Phase Evolution of the Milky Way', ApJ 972, 112, DOI 10.3847/1538-4357/ad5b60, arXiv:2310.13050, https://arxiv.org/abs/2310.13050
  2. Myeong, G. C., Belokurov, V., Aguado, D. S., Evans, N. Wyn, Caldwell, N., & Bradley, J. 2022, 'Milky Way's eccentric constituents with Gaia, APOGEE & GALAH', ApJ 938, 21, DOI 10.3847/1538-4357/ac8d68, arXiv:2206.07744, https://arxiv.org/abs/2206.07744
  3. Xiang, M., & Rix, H.-W. 2022, 'A time-resolved picture of our Milky Way's early formation history', Nature 603, 599–603, DOI 10.1038/s41586-022-04496-5, https://doi.org/10.1038/s41586-022-04496-5
  4. Bonaca, A., Conroy, C., Cargile, P. A., Naidu, R. P., Johnson, B. D., Zaritsky, D., Ting, Y.-S., Caldwell, N., Han, J. J., & van Dokkum, P. 2020, 'Timing the Early Assembly of the Milky Way with the H3 Survey', ApJL 897, L18, DOI 10.3847/2041-8213/ab9caa, https://doi.org/10.3847/2041-8213/ab9caa
  5. Nidever, D. L., Gilbert, K., Tollerud, E., Siders, C., Escala, I., Prieto, C. Allende, Smith, V., Cunha, K., Debattista, V. P., Ting, Y.-S., & Kirby, E. N. 2023, 'The Prevalence of the α-bimodality: First JWST α-abundance Results in M31', arXiv:2306.04688, https://arxiv.org/abs/2306.04688
  6. van de Sande, J., Fraser-McKelvie, A., Fisher, D. B., Martig, M., Hayden, M. R., & the GECKOS Survey 2023, 'GECKOS: Turning galaxy evolution on its side with deep observations of edge-on galaxies', arXiv:2306.00059, https://arxiv.org/abs/2306.00059
  7. Beane, A., Johnson, J. W., Semenov, V. A., Hernquist, L., Chandra, V., & Conroy, C. 2025, 'Rising from the Ashes. II. The Bar-driven Abundance Bimodality of the Milky Way', ApJ 985, 221, DOI 10.3847/1538-4357/adceab, https://doi.org/10.3847/1538-4357/adceab
  8. Amarante, J. A. S., Laporte, C. F. P., Debattista, V. P., Beraldo e Silva, L., Limberg, G., Perottoni, H. D., Li, Z.-Y., Borbolato, L., Fiteni, K., Cao, C., Deg, N., Khachaturyants, T., & Liao, X. 2026, 'GASTRO library II: Exploring Chemical Bimodalities in Disk Galaxies with GSE-like Mergers and Massive Star-forming Clumps', ApJ 1004, 242, DOI 10.3847/1538-4357/ae70fe, arXiv:2512.01293, https://arxiv.org/abs/2512.01293
  9. Feltzing, S., Feuillet, D., & Bensby, T. 2026, 'Dawn of the Milky Way disk: Determination of when a rotationally supported disk appears and dating the spin-up of the disk', ApJL 1004, L28, DOI 10.3847/2041-8213/ae6f14, arXiv:2605.20959, https://arxiv.org/abs/2605.20959
  10. Kisku, S., Schiavon, R. P., Font, A. S., Mason, A., Horta, D., Taylor, D. J., Sante, A., Fernández-Trincado, J. G., & Beers, T. C. 2025, 'Measuring and modelling the Splash with APOGEE/Gaia and ARTEMIS', MNRAS, DOI 10.1093/mnras/staf1075, arXiv:2507.15944, https://arxiv.org/abs/2507.15944
  11. Massari, D., Zerbinati, C., Fanelli, C., Helmi, A., Ceccarelli, E., Aguado-Agelet, F., Cassisi, S., Wempe, E., Monelli, M., Bellini, A., Callingham, T., Woudenberg, H. C., Cohen, R., Gallart, C., Pancino, E., Saracino, S., Salaris, M., & Mucciarelli, A. 2026, 'Evidence of a massive accretion event 1.8 billion years before the Gaia-Sausage-Enceladus merger', Nature Astronomy, DOI 10.1038/s41550-026-02931-5, arXiv:2601.18896, https://doi.org/10.1038/s41550-026-02931-5
  12. Skúladóttir, Á., Ernandes, H., Feuillet, D. K., Mori, A., Feltzing, S., Lucchesi, R. E. R., & Di Matteo, P. 2025, 'Evidence of Gaia Enceladus experiencing at least two passages around the Milky Way', ApJL 986, L21, DOI 10.3847/2041-8213/addc66, arXiv:2506.00409, https://arxiv.org/abs/2506.00409
  13. Semenov, V. A., Conroy, C., Smith, A., Puchwein, E., & Hernquist, L. 2025, 'How Early Could the Milky Way's Disk Form?', ApJ 990, 7, DOI 10.3847/1538-4357/addf48, arXiv:2409.18173, https://arxiv.org/abs/2409.18173