The dawn is quiet II: Gaia XP constraints on the Milky Way's proto-Galaxy from very metal-poor MDF tails

TL;DR

This work leverages Gaia DR3 XP metallicities to measure the slope of the Milky Way's very metal-poor MDF tail and interprets it with a grid of one-zone chemical-evolution models. The results indicate a near-unity exponential tail (k ~ 1) across catalogs, pointing to a proto-Galaxy with a moderate initial gas reservoir and sustained, low-to-moderate inflow and star formation during the first Gyr. Auriga Milky Way analogs show similar MDF slopes, reinforcing a picture of quiet, gas-regulated early growth rather than an early, massive starburst. The metal-poor MDF tail thus provides a quantitative, complementary constraint on the Milky Way’s early gas accretion and star-formation history, alongside the initial [$\alpha$/Fe] upturn, and supports a unified view of the proto-Galaxy as a well-mixed, slowly assembling system.

Abstract

The earliest phase of the Milky Way's evolution involved a transition from a dispersion-supported proto-galaxy to a rotationally supported disk. A key chemical signature of this transition is the moderate rise in [$α$/Fe] near $\mathrm{[Fe/H]}\approx-1.3$, which we previously interpreted as evidence for $α$-enhanced gas accretion fueling early disk formation. However, this trend alone does not uniquely constrain the trade-off between initial gas mass, inflow rate, and star formation efficiency (SFE), leaving the physical condition of the proto-Milky Way uncertain. To break this degeneracy, we analyze the metal-poor tail ($-3<\mathrm{[Fe/H]}<-2$) of the Milky Way's metallicity distribution function (MDF) using Gaia DR3 BP/RP (XP) metallicities from ten catalogs. After applying recommended quality cuts, all catalogs exhibit a single-slope exponential tail with slopes $k\simeq0.5$--2.0. Comparison with one-zone galactic chemical-evolution (GCE) models that replicated the [$α$/Fe]-rise from Paper I shows that shallow tails ($k\simeq0.6$) require a massive initial cold gas reservoir ($\gtrsim10^9\, \mathrm{M_\odot}$), while steeper tails ($k\gtrsim1$) arise from small reservoirs that built up over time with weak inflow. MDFs with $k \simeq 1.0$ are best reproduced under our GCE framework, which favor a proto-Galaxy with a moderate gas reservoir ($10^{8}$--$10^{9}\, \mathrm{M_\odot}$) sustained through weak continuous inflow ($\sim 2 \ \mathrm{M_\odot\,yr^{-1}}$) and SFE comparable to today's value (a few $\times 10^{-10}\,\mathrm{yr^{-1}}$) during the first Gyr. This scenario is reinforced by MDFs of 30 Milky Way analogs in the Auriga simulations, which exhibit similar slopes ($k\approx1.2$). The metal-poor MDF tail thus provides a quantitative constraint on the Milky Way's early gas accretion and star formation history.

Figures (12)

• Figure 1: Stellar density distributions in the Gaia CMD. Panels in (a) display densities based on apparent magnitude, while panels in (b) show densities using observed absolute magnitudes derived from Gaia parallaxes. Data are sourced from various Gaia XP metallicity catalogs following their quality cuts listed in Table \ref{['tab:catalog_summary']}. The quality cuts restrict several catalogs to mostly giants due to training pipelines and output quality control.
• Figure 2: Stellar distribution in Galactocentric rotational velocity versus metallicity for various Gaia XP metallicity catalogs. The red dashed line in every panel indicates the median velocity across the metallicity range. Most metal-poor stars have low rotational velocities as expected for the proto-Galaxy, but false metal-poor stars that likely belong to the disk become common as [Fe/H] exceeds -2.0. Similar contamination also appears in 2024MNRAS.531.2126F and 2025ApJS..279....7Y as [Fe/H] drops below -2.6.
• Figure 3: Stellar distribution in Galactocentric radius ($R$) versus height above the Galactic plane ($|\mathrm{z}|$) for the Gaia XP metallicity catalogs. Differences in training sets and quality controls produce distinct spatial footprints. Catalogs restricted to giants extend to larger $R$ and $|\mathrm{z}|$, whereas others more closely follow the native XP selection function as being concentrated in the solar neighborhood. A notable outcome of Gaia XP is the recovery of large numbers of metal-poor stars at low $|\mathrm{z}|$, an area that was sparsely sampled before Gaia XP.
• Figure 4: Gaia CMD colored by median Galactocentric rotational velocity ($v_\phi$) in various Gaia XP metallicity catalogs. The magnitudes are observed absolute magnitudes as in b) in Fig. \ref{['fig:gaia_xp_cmd']}. The slow-rotating stars highlight a PARSEC 1.2S isochrone for a stellar population 13 Gyr old and with [M/H] = -2.2. Most stars with radial velocities in the various catalogs also tend to reside along such an isochrone on the CMD as seen in Figure \ref{['fig:gaia_xp_cmd']}. This suggests that at least most stars with low $v_\phi$ in Gaia RVS are genuinely old metal-poor stars.
• Figure 5: Logarithmic metal-poor MDFs from various Gaia XP catalogs after applying the quality cuts listed in Table \ref{['tab:catalog_summary']}. Dashed lines show exponential density profiles with reference slopes $k = 0.6$, 1.0, and 1.4 for comparison. Despite differences among the catalogs, most exhibit a near-exponential profile with $k \approx 1.0$. Catalogs from 2024AA...691A..98K and 2025AA...695A..75Y, trained solely on APOGEE, under-represent VMP stars.