Portrait of a Galaxy on FIRE: Is the $α$-bimodality a natural consequence of inside-out disc growth in a hierarchical formation scenario?

TL;DR

The paper demonstrates that an $\alpha$-bimodality in a MW-mass disc can emerge from inside-out growth within a hierarchical assembly, without major mergers or significant radial migration. By analyzing a FIRE-2 zoom-in galaxy (Romeo) and decomposing its disc with Gaussian Mixture Modelling, the authors identify three coexisting discs—high-$\alpha$, bridge, and low-$\alpha$—that trace distinct formation epochs and radial extents. They show that gas flows, including accretion from the hot corona and minor-merger–driven inflows, drive the three-phase chemical evolution, with corona dilution and modest SFR fluctuations enabling the separation of the $\alpha$ sequences. The work highlights the importance of corona–disc interactions and non-instantaneous gas mixing for interpreting disc chemistry and cautions that radial migration is not a prerequisite for bimodality, offering a pathway to understanding disc chemo-dynamics in a cosmological context.

Abstract

The chemical dichotomy in the [$α$/Fe]-[Fe/H] plane is a consequence of the complex processes underlying the formation and evolution of disc galaxies such as observed in the stellar Milky Way disc. We determine what can drive an $α$-bimodality of the disc in a zoom-in hydrodynamical simulated galaxy which has had no major mergers and negligible radial migration. Using a Milky Way-mass galaxy from the FIRE-2 suite of simulations, we analyse gas flows in the disc together with its star formation and merger history, as well as the chemical evolution of the hot corona, to investigate their connection to transitions in the chemo-dynamical structure of the stellar disc and its radial distribution. The simulated galaxy exhibits high and low-$α$ sequences without having experienced major mergers nor significant radial migration. A high-$α$ thick disc forms during the early chaotic clustering phase. Afterwards, as the star formation rate declines, a dip in the stellar number density appears, coinciding with the dilution of the galactic corona by a minor merger, which subsequently halts the rise of [Fe/H] in the disc. Later, accreted gas onto the disc from minor mergers, mildly enhances the star formation rate and generates the low-$α$ sequence in the outer disc, with radial inward flows of this material feeding the low-$α$ inner disc. Furthermore, we find that even at fixed radii, newly formed stars retain a sizable spread in their chemical abundances, reflecting chemical differences between the in-situ and the infalling gas from which they formed, further indicating that instantaneous gas mixing is invalid. Understanding the chemical evolution of stellar discs requires accounting for their accretion merger history and interaction with the surrounding hot corona, as well as the vertical and radial gas flows that redistribute metals within the disc.

Figures (15)

• Figure 1: Top: Stellar surface density profiles of the Romeo and Vintergatan 2021MNRAS.503.5826A simulated galaxies, together with the Milky Way profile taken from 2024NatAs...8.1302L and the best-fitting morphology of 2011MNRAS.414.2446M. The former Milky Way profile has been normalised to enclose a stellar mass of $4\times10^{10}\,\rm M_\odot$. The dot-dashed orange line depicts a falling exponential profile with scale-length of 2.6 kpc. Bottom: Star formation rate as a function of time in Romeo calculated with timesteps $\Delta t = 10$ Myr (orange), $\Delta t = 100$ Myr (purple) and $\Delta t = 500$ Myr (blue).
• Figure 2: One-dimensional age distribution of all disc star particles along with the disc components identified by GMM. The vertical grey lines mark key evolutionary stages of the disc identified by GMM (see Sect. \ref{['subsec:discs']}): the onset of the thick or high-$\alpha$ disc (10.9 Gyr ago), the time when the bridge or intermediate-$\alpha$ disc begins to dominate over thick disc (7.7 Gyr ago), and the onset of the $\alpha-$bimodality (3.6 Gyr ago). The vertical grey bands mark the timing of the bar episode.
• Figure 3: Two-dimensional distribution of the azimuthal or rotational velocity of all stellar particles in Romeo$R < 40{\, \rm kpc}$ and $|z| < 10{\, \rm kpc}$ as a function of age (left panel) and metallicity [Fe/H] (right panel), with the median trend shown in red. The green, blue and purple lines depict the median trend of the stellar population in the high-$\alpha$, bridge and low-$\alpha$ discs, respectively. In the left panel, the dashed vertical lines indicate the inferred period at which the Milky Way's stellar disc settles (see Sect. \ref{['subsec:discs']}), while the dashed black line in the right panel shows the inferred trend for the Milky Way as in belokurov_dawn_2022.
• Figure 4: Top panel: two-dimensional number density of a sample of giants in Gaia DR3 crossmatched with APOGEE DR17 that spanned a Galactocentric range of 4-16 kpc (see text for details). The red dashed lines delimit the thick disc/bridge/low-$\alpha$ regions used to derive the kinematic properties in Table \ref{['tab:disc']}, while the excluded region for this derivation is shaded red. Middle panel: two-dimensional number density of star disc particles, as identified by GMM, in the [Mg/Fe] vs. [Fe/H] plane. Green, blue and purple mark the contours containing 90% of the high-$\alpha$, bridge and low-$\alpha$ disc stars, respectively. Bottom panel: Mean age distribution of disc stars in the same plane, with the black contour enclosing the region containing 90% of the star particles in the disc. The last two panels show disc star particles at $z\sim0.1$.
• Figure 5: Time evolution of radial stellar surface density (top) and vertical density (bottom) profiles from $z=1.5$ to $z=0$ for the disc components identified using GMM. The thickest solid grey line represents the total disc surface density profile at $z=0$, while the dashed lines show exponential profiles with scale-lengths $r_d$, obtained by fitting the $z=0$ surface density profile of each substructure. The vertical profiles are shown in a cylindrical bin within $r_d\pm1\,\rm kpc$ (see e.g. 2021ApJS..254....2P), where $r_d$ is calculated at each lookback time.