Challenges to the Two-Infall Scenario by Large Stellar Age Catalogs

Abstract

Stars in the Milky Way disk exhibit a clear separation into two chemically distinct populations based on their [$α$/Fe] ratios. This $α$-bimodality is not a universal feature of simulated disk galaxies and may point to a unique evolutionary history. A popular and well-studied explanation is the two-infall scenario, which postulates that two periods of substantial accretion rates dominate the assembly history of the Galaxy. Thanks to recent advances in stellar age measurements, we can now compare this model to more direct measurements of the Galaxy's evolutionary timescales across the disk. We run multi-zone galactic chemical evolution models with a two-infall-driven star formation history and compare the results against abundance patterns from APOGEE DR17, supplemented with stellar ages estimated through multiple methods. Although the two-infall scenario offers a natural explanation for the $α$-bimodality, it struggles to explain several features of the age--abundance structure in the disk. First, our models generically require a massive and long-lasting dilution event, but the data show that stellar metallicity is remarkably constant across much of the lifetime of the disk. This apparent age-independence places considerable restrictions upon the two-infall parameter space. Second, most local metal-rich stars in APOGEE have intermediate ages, yet our models predict these stars should either be very old or very young. Some of these issues can be mitigated, but not completely resolved, by pre-enriching the accreted gas to low metallicity. These restrictions also place limits on the role of merger events in shaping the chemical evolution of the thin disk.

Figures (11)

• Figure 1: Chemical abundance tracks predicted by a two-infall model at $R_{\rm gal}=8\,{\rm kpc}$ (solid curve) versus a model with a smooth SFH at three different radii (dashed curves). The two-infall model adopts the fiducial parameters according to Table \ref{['tab:parameters']}, while the smooth SFH adopts the parameters of the "inside-out" model of johnson_stellar_2021. Both models assume the $y/Z_\odot=2$ yield scale (see Table \ref{['tab:yields']}). The 2-D histogram shows the number of stars from APOGEE DR17 in the Solar neighborhood ($7\leq R_{\rm gal}<9\,{\rm kpc}$, $0\leq|z|<2\,{\rm kpc}$) in each bin of ([Fe/H], [O/Fe]).
• Figure 2: Comparison between NN leung_variational_2023 and [C/N] roberts_cn_2025 ages for our sample. Left: Comparison of age estimates from both methods for all APOGEE stars. The solid black curve plots the rolling median (window size 1000 stars) of the [C/N] age estimate as a function of the NN age, the dashed line indicates the one-to-one correspondence, and the white error bars indicate the age uncertainty as a function of NN age. Right: The local ($7\leq R_{\rm gal}<9\,{\rm kpc}$, $|z|<0.5\,{\rm kpc}$) age--metallicity relation for each age estimation method. The black curve plots the rolling median [Fe/H] as a function of age.
• Figure 3: The evolution of (a) the infall surface density, (b) the star formation surface density, (c) the gas surface density, and (d) the star formation efficiency timescale as a function of time for our fiducial multi-zone model with $y/Z_\odot=1$. Each panel plots the evolution in six different zones of width $\delta R_{\rm gal}=0.1\,{\rm kpc}$, color-coded by Galactocentric radius.
• Figure 4: The ISM abundance evolution at $R_{\rm gal}=8\,{\rm kpc}$ of three multi-zone models at different yield scales (see Table \ref{['tab:models']}), each of which feature a major dilution event $10-11\,{\rm Gyr}$ ago. The 2-D histograms show the APOGEE stellar age--abundance distributions in the Solar neighborhood ($7\leq R_{\rm gal}<9\,{\rm kpc}$, $|z|<0.5\,{\rm kpc}$), adopting the leung_variational_2023 ages. The black curve (points) plots the rolling median (binned mode) of the abundance data as a function of age, and the gray error bars along the bottom of each panel indicate the median age and abundance errors as a function of age. The gray dashed curve plots the abundance evolution of the palicio_analytic_2023 analytic model, which our yZ2-earlyonset model (green curve) closely matches. The left-hand marginal panels show the predicted (blue, pink, and green) and observed (gray) stellar abundance distributions, which are boxcar-smoothed with a width of 0.05 dex for visual clarity.
• Figure 5: Stellar age--abundance relations predicted by multi-zone models at the $y/Z_\odot=1$ yield scale (see Table \ref{['tab:yields']}). Each point represents a stellar population drawn from the Solar neighborhood near the midplane ($7\leq R_{\rm gal}< 9\,{\rm kpc}$, $|z| < 0.5\,{\rm kpc}$) color-coded by its birth radius. In this and subsequent figures, a Gaussian scatter is applied to each point according to the median age and abundance uncertainties in Table \ref{['tab:uncertainties']}. For visual clarity, we plot only a random mass-weighted sample of 10000.0 points in each panel. The dashed curve plots the predicted ISM abundance at $R_{\rm gal}=8\,{\rm kpc}$, and the solid black curve plots the rolling median stellar abundance. The red curve plots the rolling median abundance of the APOGEE sample, and the shaded regions are the 16th--84th percentile ranges. Each column shows results from a different multi-zone model: (a) our fiducial model, yZ1-fiducial; (b) a model with stronger radial migration, yZ1-migration; (c) a model with a higher local thick-to-thin disk ratio, yZ1-diskratio; and (d) a model with pre-enriched gas infall, yZ1-preenrich (see Table \ref{['tab:models']} for details).