A time-resolved picture of our Milky Way’s early formation history

Understanding the Milky Way’s assembly requires knowing when stars formed, from what chemical material, and in what dynamical components (halo versus disks). Achieving this demands precise stellar ages reaching back to the earliest epochs (~14 Gyr). Subgiant stars offer particularly accurate ages because their luminosities are strongly age-sensitive during a brief evolutionary phase and their surface abundances reflect the composition of their birth gas. Large spectroscopic and astrometric surveys now enable the construction of a vast, well-characterized subgiant sample to decode the Galaxy’s time-resolved age–metallicity distribution p(τ, [Fe/H]) and thereby reconstruct the formation history of the halo and disk components. This study aims to provide a quantitative, time-resolved picture of the Milky Way’s early formation and chemical enrichment history using such a sample.

Prior work has qualitatively separated the Milky Way into halo and disk components and suggested multiple assembly phases, including an early, α-enhanced thick disk and later thin-disk growth, as well as evidence for significant merger events such as the Gaia-Sausage-Enceladus (GSE). Earlier studies hinted at age–metallicity trends and possible overlaps between halo and thick-disk formation but often lacked the sample size, spatial coverage, or age precision to draw definitive, time-resolved conclusions. Cosmological simulations (e.g., IllustrisTNG) predict multiple star-formation episodes triggered by mergers and interactions, with quiescent phases in between. Observationally, precise ages across large samples are key to testing these predictions; subgiant-based analyses offer improved age precision over main-sequence turn-off stars, mitigating issues such as atomic diffusion that can bias abundance interpretations.

Data sources and sample construction: Using Gaia eDR3 astrometry and photometry together with LAMOST DR7 spectroscopy, the authors identified approximately 247,104–250,000 subgiant stars based on their positions in the effective temperature–absolute magnitude diagram. The sample spans Galactocentric radii ~6–14 kpc, ages ~1.5–13.8 Gyr, and metallicities −2.5 < [Fe/H] < 0.4. A straightforward selection function permits estimation of space densities and corrections for volume selection effects.

Age and abundance determination: Stellar ages τ were inferred by Bayesian fitting to Yonsei–Yale (YY) isochrones, combining Gaia parallaxes (distances) and apparent magnitudes with spectroscopic [Fe/H], [α/Fe] (e.g., Mg, Si, Ca, Ti, Zr) and atmospheric parameters (Teff, log g). The resulting median relative age uncertainty is ~7.5% across 1.5 Gyr to ~13.3 Gyr.

Kinematic and chemical stratification: To disentangle Galactic components, the analysis divided stars by angular momentum Jφ (Lz) and α-enhancement [α/Fe], without directly selecting on τ or [Fe/H]. High-angular-momentum, low-[α/Fe] stars trace the dynamically cold disk; low-angular-momentum and α-enhanced stars trace older, hotter components (old thick disk and halo). Distributions considered include p(τ, [Fe/H]), p([Fe/H]) and its morphology (e.g., V-shape indicative of radial migration), and volume-corrected p(τ, [Fe/H]) to infer relative star-formation histories. The authors also examined subsets split by orbital characteristics to separate predominantly halo-like from disk-like stars and investigated retrograde/low-angular-momentum subpopulations associated with the GSE merger.

Modeling and diagnostics: The tightness of the age–metallicity relation along the old disk sequence was quantified; a simple model yields an intrinsic age dispersion <0.2 Gyr at fixed [Fe/H] over a ~6 Gyr interval, implying well-mixed star-forming gas. Conversely, the younger disk shows larger [Fe/H] dispersion at fixed age, consistent with metallicity gradients and radial migration in a quiescent disk phase.

• The stellar age–metallicity distribution p(τ, [Fe/H]) bifurcates into two nearly disjoint components separated at τ ≈ 6 Gyr.
• Younger sequence: reflects a late, dynamically quiescent phase of disk formation with strong signatures of radial migration, producing a V-shaped p([Fe/H]) morphology tied to the Galactic radial metallicity gradient and migration from inner and outer disk regions.
• Older sequence: comprises the stellar halo and an old, α-enhanced (thick) disk. The old thick disk began forming ~13 Gyr ago (~0.8 Gyr after the Big Bang), about 2 Gyr earlier than the final assembly of the inner halo.
• Most stars in the old thick disk formed around ~11 Gyr ago, coincident with the Gaia-Sausage-Enceladus (GSE) merger; the star-formation rate in the old disk peaked at ~11.2 Gyr and then declined.
• The old disk’s ISM underwent continuous enrichment by >1 dex in [Fe/H] (from ~−1.0 to ~+0.5) over ~5–6 Gyr, while remaining well mixed: the intrinsic scatter is <0.2 Gyr in age at fixed [Fe/H], and the [Fe/H] dispersion at a given age is <~0.2 dex across a wide metallicity range.
• At [Fe/H] ≈ −1, the old disk sequence is ~2 Gyr older than the halo sequence, producing a Z-shaped structure in p(τ | [Fe/H]).
• A population of low-angular-momentum “splashed” old-disk stars associated with the GSE event supports that the major merger was largely complete ~11 Gyr ago, somewhat earlier than some previous estimates (~10 Gyr).
• The findings collectively indicate multiple assembly phases: early rapid formation of an α-enhanced thick disk overlapping with halo assembly, followed by a later, quiescent thin-disk growth with substantial radial migration.

By leveraging precise subgiant ages over a large, well-characterized sample, the study reconstructs a time-resolved formation sequence of the Milky Way. The observed bifurcation at ~6 Gyr cleanly separates an early, turbulent phase (halo plus α-enhanced thick disk) from a later, secular disk growth phase. The tight age–metallicity relation of the old disk implies a thoroughly mixed ISM during its ~5–6 Gyr formation epoch, contrasting with the larger metallicity dispersion in the younger disk where radial gradients and migration dominate. The star-formation rate peak at ~11.2 Gyr aligns with the completion of the GSE merger, supporting a scenario in which mergers trigger enhanced star formation and shape the kinematic structure (including splashed, low-angular-momentum stars). The ~2 Gyr age offset between old disk and halo sequences at [Fe/H] ≈ −1 clarifies the relative timing of disk and halo assembly. These observations provide strong empirical constraints for chemo-dynamical models and are broadly consistent with cosmological simulations that predict merger-induced star formation episodes separated by quieter phases.

This work delivers a quantitative, time-resolved picture of the Milky Way’s early formation by exploiting precise ages of ~250,000 subgiant stars from Gaia eDR3 and LAMOST DR7. The Galaxy’s old, α-enhanced thick disk began forming ~13 Gyr ago, experienced sustained, well-mixed chemical enrichment over 5–6 Gyr, and reached a star-formation peak around ~11.2 Gyr coincident with the GSE merger; the inner halo’s final assembly occurred ~2 Gyr later than the thick-disk onset. A later, dynamically quiescent disk phase shows clear signatures of radial migration. Together, these results delineate distinct assembly phases and set stringent constraints for Galactic chemo-dynamical evolution.

Potential future directions include: extending subgiant-based age dating to larger Galactic volumes (inner/outer disk and bulge), incorporating broader elemental abundance dimensions (multi-element chemistry) to refine mixing and enrichment constraints, combining with full dynamical modeling and upcoming Gaia releases for improved orbital characterization, and cross-validating with other spectroscopic surveys to better control selection effects and systematics.

• Sample limitations: Subgiants are intrinsically rare; the selection imposes a practical lower age limit (~1.5 Gyr) to avoid contamination from other evolutionary phases (e.g., horizontal branch), excluding the very youngest populations.
• Spatial coverage: The sample primarily spans Galactocentric radii ~6–14 kpc; very inner Galaxy and far outer halo regions are underrepresented.
• Systematics and modeling: Age estimates depend on isochrone models (YY) and assumed stellar parameters; residual systematics in parallaxes, photometry, or spectroscopic abundances (including [α/Fe]) may affect inferred ages and metallicities.
• Selection and volume corrections: Results rely on a characterized selection function and volume corrections; any residual biases could influence star-formation history inferences.
• Kinematic partitioning: Component separation using angular momentum and [α/Fe] thresholds may misclassify some stars in overlap regions, potentially blurring boundaries between halo and disk sequences.
• Chemical assumptions: Inferences about ISM mixing from tight age–metallicity relations presume minimal unresolved abundance systematics and negligible unmodeled gradients during the thick-disk epoch.