Gaia Sausage-Enceladus: Milky Way’s Ancient Merger

• Gaia Sausage-Enceladus (GSE) is an ancient merger remnant identified by its narrow metallicity distribution and distinct chemodynamical properties.
• It is characterized through a combination of astrometric and spectroscopic data that trace its highly radial, low-angular momentum stellar orbits across the inner and outer halo.
• Studies of GSE reveal a complex star formation and chemical evolution history that has influenced the Milky Way’s halo structure, disk chemistry, and even the Galactic warp.

Gaia-Sausage-Enceladus (GSE), also written Gaia-Sausage/Enceladus or GS/E, is the dominant ancient accretion remnant in the Milky Way stellar halo and is commonly interpreted as the debris of the Galaxy’s last major merger, which occurred about $10$ Gyr ago. In the Gaia-era picture, GSE is identified as a chemically coherent, accreted population on highly radial, low-$L_z$ orbits that dominates the nearby accreted halo and much of the inner halo. Recent work has extended GSE from a local phase-space feature into a multi-scale fossil record: it is now studied through global halo morphology, distant shells and retrograde streams out to $100$ kpc, detailed abundance clocks, globular-cluster age-metallicity relations, and cosmological analogues that test whether the remnant is monolithic or composite (Feuillet et al., 2021, Chandra et al., 2022).

1. Definition and observational identification

GSE is observationally defined through the conjunction of dynamics and chemistry. In Gaia and APOGEE-based selections, its stars occupy a highly radial, low-$L_z$ region of action or integral-of-motion space and are chemically distinct from in-situ thick-disk contamination. A widely used action-space selection is

$$30 \le \sqrt{J_R} \le 55\,(\mathrm{kpc\,km\,s^{-1}})^{1/2}, \qquad -500 \le L_z \le 500\ \mathrm{kpc\,km\,s^{-1}},$$

which was adopted as the main GSE selection in APOGEE + Gaia analyses; a narrower Myeong et al. action selection overlaps with this region but yields a mean $[\mathrm{Fe/H}]$ about $0.15$ dex higher, indicating stronger contamination by metal-richer stars (Feuillet et al., 2021).

Within such selections, GSE has a narrow metallicity distribution function centered near $[\mathrm{Fe/H}] \sim -1.2$. Published characterizations include a mean $[\mathrm{Fe/H}] \sim -1.15$, a chemically cleaned mean/median of $-1.23/-1.20$ dex, and, in a self-consistent APOGEE DR17 + Gaia EDR3 analysis, a median $L_z$0 of $L_z$1 dex with $L_z$2 dex dispersion (Feuillet et al., 2021, Limberg et al., 2022). In abundance space, GSE follows a single $L_z$3 sequence with an $L_z$4-knee at $L_z$5, and it is characterized by uniformly low $L_z$6 to $L_z$7 dex; this low-Al locus is one of the clearest chemical signatures of accreted origin in APOGEE analyses (Feuillet et al., 2021).

Age estimates from nearby Gaia-based samples place the dominant GSE population at $L_z$8–$L_z$9 Gyr, consistent with a merger at $100$0 and with the interpretation of GSE as the Milky Way’s last major accretion event (Feuillet et al., 2021). This identification framework is now standard, but the literature also emphasizes that dynamics alone are insufficient: chemical cleaning is necessary to remove heated thick-disk and in-situ halo contaminants, especially at the metal-rich end (Feuillet et al., 2021).

2. Chemodynamical structure in the inner halo

The canonical dynamical description of GSE is a radially anisotropic halo population with low net rotation and high eccentricity. A chemically selected APOGEE DR17 sample shows that its anisotropy profile is not constant with radius: it remains high and nearly constant with $100$1 beyond $100$2 kpc, but drops inward to $100$3 at $100$4–$100$5 kpc, with

$100$6

A two-component Osipkov-Merritt model with $100$7 kpc, $100$8 kpc, and $100$9 fits these data better than a constant-anisotropy model, implying that the inner remnant is substantially less radial than the outer remnant (Lane et al., 4 Sep 2025).

Inner-halo overdensities long discussed as separate structures are now widely connected to GSE. A combined SEGUE, APOGEE, Gaia, and StarHorse analysis of the Hercules-Aquila Cloud (HAC) and Virgo Overdensity (VOD) found that these overdensities are dominated by highly radial orbits, with about $L_z$0 of VOD stars and $L_z$1 of HAC stars having $L_z$2. In the $L_z$3–$L_z$4 and $L_z$5–$L_z$6 planes, HAC and VOD occupy the same low-$L_z$7, high-eccentricity region as the prototypical GSE population, and their metallicity distribution functions are strongly peaked and statistically compatible with GSE: $L_z$8 for HAC-S, $L_z$9 for HAC-N, $$30 \le \sqrt{J_R} \le 55\,(\mathrm{kpc\,km\,s^{-1}})^{1/2}, \qquad -500 \le L_z \le 500\ \mathrm{kpc\,km\,s^{-1}},$$0 for VOD, and $$30 \le \sqrt{J_R} \le 55\,(\mathrm{kpc\,km\,s^{-1}})^{1/2}, \qquad -500 \le L_z \le 500\ \mathrm{kpc\,km\,s^{-1}},$$1 for GSE (Perottoni et al., 2022).

APOGEE abundance comparisons reinforce this identification. In $$30 \le \sqrt{J_R} \le 55\,(\mathrm{kpc\,km\,s^{-1}})^{1/2}, \qquad -500 \le L_z \le 500\ \mathrm{kpc\,km\,s^{-1}},$$2–$$30 \le \sqrt{J_R} \le 55\,(\mathrm{kpc\,km\,s^{-1}})^{1/2}, \qquad -500 \le L_z \le 500\ \mathrm{kpc\,km\,s^{-1}},$$3, $$30 \le \sqrt{J_R} \le 55\,(\mathrm{kpc\,km\,s^{-1}})^{1/2}, \qquad -500 \le L_z \le 500\ \mathrm{kpc\,km\,s^{-1}},$$4–$$30 \le \sqrt{J_R} \le 55\,(\mathrm{kpc\,km\,s^{-1}})^{1/2}, \qquad -500 \le L_z \le 500\ \mathrm{kpc\,km\,s^{-1}},$$5, $$30 \le \sqrt{J_R} \le 55\,(\mathrm{kpc\,km\,s^{-1}})^{1/2}, \qquad -500 \le L_z \le 500\ \mathrm{kpc\,km\,s^{-1}},$$6–$$30 \le \sqrt{J_R} \le 55\,(\mathrm{kpc\,km\,s^{-1}})^{1/2}, \qquad -500 \le L_z \le 500\ \mathrm{kpc\,km\,s^{-1}},$$7, and $$30 \le \sqrt{J_R} \le 55\,(\mathrm{kpc\,km\,s^{-1}})^{1/2}, \qquad -500 \le L_z \le 500\ \mathrm{kpc\,km\,s^{-1}},$$8–$$30 \le \sqrt{J_R} \le 55\,(\mathrm{kpc\,km\,s^{-1}})^{1/2}, \qquad -500 \le L_z \le 500\ \mathrm{kpc\,km\,s^{-1}},$$9, HAC-S, HAC-N, and VOD overlap the GSE contours and lie in the accreted-stellar-population region, with only about $[\mathrm{Fe/H}]$0 of stars in the in-situ locus defined by $[\mathrm{Fe/H}]$1–$[\mathrm{Fe/H}]$2 criteria (Perottoni et al., 2022). A plausible implication is that several prominent inner-halo overdensities are not independent progenitors, but spatially unmixed portions of the same GSE debris field.

3. Global morphology and distant debris

Orbit-based reconstructions from local metal-poor stars show that GSE is not merely a local phase-space clump but a global halo component with a characteristic spatial scale. For stars with GSE-like kinematics and $[\mathrm{Fe/H}]$3, the reconstructed density is more spherical than the general halo, with $[\mathrm{Fe/H}]$4, and it shows an outer ridge at $[\mathrm{Fe/H}]$5 kpc. That ridge has been interpreted as an apocenter pile-up or as the characteristic radial boundary of the main debris cloud, consistent with a progenitor orbit whose apocenter was $[\mathrm{Fe/H}]$6 kpc (Sato et al., 2021).

Other density reconstructions recover a more elongated morphology. DF-corrected fits to high-purity APOGEE selections find a shallow inner density profile with a break between $[\mathrm{Fe/H}]$7 and $[\mathrm{Fe/H}]$8 kpc, a triaxial nearly prolate shape with axis ratios $[\mathrm{Fe/H}]$9, and a major axis oriented about $0.15$0 from the Sun–Galactic centre line and $0.15$1 above the plane (Lane et al., 2023). Tailored H3-based $0.15$2-body simulations of a retrograde GSE merger reproduce an elongated triaxial inner halo with axis ratios $0.15$3, whose major axis is at $0.15$4 to the plane and connects GSE apocenters; the same model explains HAC and VOD as apocenter pile-ups and predicts a double-break density profile with breaks at $0.15$5–$0.15$6 kpc and $0.15$7 kpc (Naidu et al., 2021).

The influence of GSE on halo shape has also been isolated statistically. A Gaussian-mixture decomposition of 11,624 LAMOST K giants separates a GSE-related halo from a GSE-removed halo and finds that both are vertically flattened at small radii and rounder at large radii, but that the GSE-related halo is systematically less flattened. Using

$0.15$8

the inferred slopes are $0.15$9 for the GSE-related halo and $[\mathrm{Fe/H}] \sim -1.2$0 for the GSE-removed halo; after bias correction, the intrinsic difference in $[\mathrm{Fe/H}] \sim -1.2$1 is thought to be within about $[\mathrm{Fe/H}] \sim -1.2$2 at most radii (Wu et al., 2022).

The outer halo now provides direct evidence for distant GSE structure. An all-sky Gaia DR3 XP analysis of luminous red giant stars maps the outer halo with kinematics and metallicities out to $[\mathrm{Fe/H}] \sim -1.2$3 kpc and identifies strong overdensities, including the previously identified Outer Virgo Overdensity, that are relatively metal-rich and on predominantly retrograde orbits. These are argued to be apocentric shells of GSE debris, forming $[\mathrm{Fe/H}] \sim -1.2$4–$[\mathrm{Fe/H}] \sim -1.2$5 kpc counterparts to the $[\mathrm{Fe/H}] \sim -1.2$6–$[\mathrm{Fe/H}] \sim -1.2$7 kpc shells that dominate the inner stellar halo. The same work also finds evidence for a coherent stream of retrograde stars encircling the Milky Way from $[\mathrm{Fe/H}] \sim -1.2$8–$[\mathrm{Fe/H}] \sim -1.2$9 kpc, in the same plane as the Sagittarius stream but moving in the opposite direction (Chandra et al., 2022). This extends GSE from an inner-halo remnant to a Galaxy-scale debris system.

4. Progenitor star formation and chemical evolution

Detailed abundance work has recast GSE as a resolved dwarf-galaxy chemical-evolution problem. Using SAGA abundances combined with Gaia DR3 astrometry and radial velocities, a 73-star GSE subset selected with the Feuillet et al. kinematic criterion was analyzed in $[\mathrm{Fe/H}] \sim -1.15$0, $[\mathrm{Fe/H}] \sim -1.15$1, $[\mathrm{Fe/H}] \sim -1.15$2, and $[\mathrm{Fe/H}] \sim -1.15$3. The resulting picture is that GSE underwent prolonged, inefficient star formation lasting over $[\mathrm{Fe/H}] \sim -1.15$4 Gyr, followed by abrupt quenching when it merged with the Milky Way. The paper states that “The increase in [Eu/Mg] suggests star formation lasted beyond $[\mathrm{Fe/H}] \sim -1.15$5 Gyr,” and that GSE was “quenched at $[\mathrm{Fe/H}] \sim -1.15$6, preventing a transition to a regime dominated by Ba enrichment from AGB stars” (Ernandes et al., 10 May 2025).

ChronoGal CMD fitting of Gaia DR3 stars near the Sun yields a compatible but more resolved chronology. Across three dynamical GSE selections, the deSFH shows an almost linear age–metallicity relation with three old populations and a smaller fourth one. The three oldest populations imply bulk star formation lasting for $[\mathrm{Fe/H}] \sim -1.15$7–$[\mathrm{Fe/H}] \sim -1.15$8 Gyr, beginning at about $[\mathrm{Fe/H}] \sim -1.15$9–$-1.23/-1.20$0 Gyr ago and ending about $-1.23/-1.20$1 Gyr ago, with metallicities ranging from $-1.23/-1.20$2 to $-1.23/-1.20$3. The younger, more metal-rich population at $-1.23/-1.20$4 Gyr and $-1.23/-1.20$5 is explicitly treated as uncertain in its link to GSE (González-Koda et al., 27 Feb 2025). This suggests that nearby GSE samples preserve at least two major star-formation epochs and possibly a small, more metal-rich tail whose origin remains unresolved.

The present-day debris also preserves information about the progenitor’s internal metallicity structure. A probabilistic APOGEE DR17 decomposition of local halo stars within $-1.23/-1.20$6 kpc of the Sun recovers a weak but systematic negative debris metallicity gradient of approximately $-1.23/-1.20$7 dex/kpc, interpreted as the scrambled remnant of the progenitor’s original radial abundance gradient. Calibrated against HESTIA and idealized $-1.23/-1.20$8-body mergers, this implies a progenitor radial metallicity gradient of approximately $-1.23/-1.20$9 dex/kpc (Khoperskov et al., 2023).

Lithium adds an independent chemical-evolution constraint. GSE members selected from GALAH DR3, Gaia-ESO, and SAGA with the Feuillet et al. $L_z$00–$L_z$01 criterion show both the classic main-sequence Spite plateau and an early red-giant-branch Li plateau. For $L_z$02 eRGB stars from GALAH DR3, the mean observed plateau at $L_z$03 is $L_z$04 dex with dispersion $L_z$05 dex; the main-sequence sample gives a mean $L_z$06 dex with standard deviation $L_z$07 dex at low metallicity (Nguyen et al., 22 May 2026). These results reinforce the view that GSE was chemically evolving slowly enough to preserve both standard metal-poor Li plateaus and truncated high-metallicity enrichment.

5. Globular clusters and stellar fossils

GSE is also reconstructed through its associated globular-cluster system. A combined APOGEE DR17 + Gaia EDR3 analysis of stars and globular clusters found no evidence for atypical $L_z$08 spreads among probable GSE globular clusters with APOGEE observations, except for $L_z$09 Centauri. Under the assumption that $L_z$10 Cen is a stripped nuclear star cluster, the inferred progenitor stellar mass is $L_z$11, leading to the proposal that GSE is the best available candidate for the original host galaxy of $L_z$12 Cen (Limberg et al., 2022).

Homogeneous HST-based ages for 13 dynamically associated GSE globular clusters refine this picture. The majority form a tight age-metallicity relation, but NGC 288 and NGC 6205 are more than $L_z$13 Gyr older than other GSE clusters at similar metallicity and are therefore interpreted as likely in-situ, while NGC 7099 is somewhat younger than the average GSE trend and NGC 5286 is mildly older. Excluding such outliers, the remaining clusters define two formation epochs, each of duration $L_z$14 Gyr and separated by $L_z$15 Gyr, centered at about $L_z$16 Gyr and $L_z$17 Gyr (Aguado-Agelet et al., 27 Feb 2025). This is in excellent agreement with the bursty field-star age-metallicity relation from ChronoGal and supports episodic star formation in the progenitor.

Rare nucleosynthetic tracers show that GSE preserves not only bulk chemical evolution but also signatures of extreme enrichment events. High-resolution Subaru/HDS spectroscopy of LAMOST J0804+5740 identified the first confirmed actinide-boost star within GSE: a very metal-poor star with $L_z$18, $L_z$19, and $L_z$20. Comparative modeling found that a magnetorotationally driven jet supernova $L_z$21-process model with $L_z$22 provides the best fit to its heavy-element pattern, while kinematic analysis suggested that about two-thirds of known actinide-boost stars are ex situ (Lin et al., 12 May 2025). A plausible implication is that accreted systems like GSE are efficient repositories of rare $L_z$23-process events that may be underrepresented in in-situ populations.

6. Mass, progenitor structure, and multiplicity

Published estimates of the GSE progenitor and remnant span a broad range because they depend strongly on sample purity, density modeling, and the mapping from stellar mass to halo mass. High-purity APOGEE DR16 selections corrected for kinematic incompleteness yield a remnant stellar mass

$L_z$24

implying that GS/E could make up as little as $L_z$25–$L_z$26 of the Milky Way stellar halo and corresponding, through standard stellar-mass-to-halo-mass relations, to a minor $L_z$27 merger at the time of accretion (Lane et al., 2023). By contrast, tailored H3 simulations favor a progenitor with $L_z$28 and $L_z$29, i.e. a $L_z$30 total mass merger that delivered $L_z$31 of the Milky Way’s present-day dark matter and $L_z$32 of its stellar halo (Naidu et al., 2021). Other work frames GSE as roughly comparable in mass to the present-day Large Magellanic Cloud and, in a broader formation-history context, as a merger with mass ratio roughly $L_z$33 (Ernandes et al., 10 May 2025, Perottoni et al., 2022). The literature therefore does not support a single settled progenitor mass scale.

The longstanding assumption that GSE is the remnant of one progenitor is also under active revision. In IllustrisTNG50, $L_z$34 Milky Way analogues were examined and $L_z$35 host GSE-like debris under the criteria $L_z$36 and $L_z$37; of these, $L_z$38 host a single RA merger and $L_z$39 host an RA pair, so about one third of GSE-like cases arise from two mergers. Single-merger cases have a median infall time $L_z$40 Gyr ago, whereas two-merger cases are earlier, with an abstract-level median of $L_z$41 Gyr ago (Folsom et al., 2024). This suggests that GSE-like inner halos are not uniquely diagnostic of a single massive accretion event.

Observationally, DESI-based unsupervised clustering has now identified four distinct substructures within the canonical GSE region: GSE-GSH1, GSE-GSH2, GSE-GSH3, and GSE-GSH4, with ages of about $L_z$42, $L_z$43, $L_z$44, and $L_z$45 Gyr, respectively. All four are broadly consistent with the overall phase-space distribution and abundance patterns of GSE, but they occupy offset loci in $L_z$46–$L_z$47, $L_z$48–$L_z$49, and action space and show chemical differences, especially in $L_z$50, $L_z$51, and $L_z$52 (Wang et al., 3 Jun 2026). A graph-attention analysis of GALAH DR4 abundance space reaches a related conclusion from a different direction: stars dynamically associated with GSE split into two chemically distinct clusters, GSE 1 and GSE 2. GSE 1 is more metal-poor and more $L_z$53-enhanced; GSE 2 is more metal-rich and less $L_z$54-enhanced; and the bootstrapped median trend rises from $L_z$55 dex at the highest energies to $L_z$56 dex at the lowest energies (Quandt-Rodriguez et al., 2 Feb 2026). These results are consistent with either internal stratification in a single massive progenitor or a genuinely composite origin assembled through multiple sequential merger episodes.

7. Consequences for Milky Way evolution

GSE is not only a halo fossil; it has been invoked as a driver of several large-scale Milky Way properties. H3 evidence for a population of highly retrograde stars with chemistry nearly identical to radial GSE debris led to the proposal that GSE entered the Galaxy on a highly retrograde orbit and was subsequently radialized by dynamical friction. In that interpretation, the highly retrograde “Arjuna” stars are outer-disk debris stripped early in the merger, whereas the radial sausage debris comes from the inner disk and was stripped later. The same simulations explain HAC and VOD as apocenter pile-ups and predict highly retrograde outer-halo streams containing about $L_z$57 of GSE stars at $L_z$58 kpc (Naidu et al., 2021). The distant shell and retrograde-stream detections in Gaia DR3 XP data provide direct observational support for this outer-halo extension (Chandra et al., 2022).

The merger trajectory has also been connected to disk chemistry. A Galactic chemical-evolution model that inserts GSE at $L_z$59 Gyr with $L_z$60 Gyr and a significantly retrograde trajectory $L_z$61 argues that radial and retrograde mergers drive strong inward gas flows, lower the gas surface density across much of the disk, and thereby accelerate the decline in $L_z$62 through continued Type Ia enrichment at reduced star-formation rate. In these models, radial or retrograde trajectories increase the low-$L_z$63 population and can produce a bimodal $L_z$64 distribution at fixed $L_z$65, provided a substantial high-$L_z$66 sequence already exists (Johnson et al., 9 Oct 2025). This suggests that GSE’s orbital geometry, not only its mass, may be relevant to the origin of the Milky Way disk’s chemical bimodality.

A further proposal links GSE to the Galactic warp. Hydrodynamical GIZMO simulations of a gas-rich GSE merger with Milky Way progenitor mass $L_z$67, GSE progenitor mass $L_z$68, GSE stellar mass $L_z$69, and GSE gas fraction $L_z$70 reproduce a long-lived, nonsteady, asymmetric warp in both stellar and gas disks, with present-day precession rates of $L_z$71 for $L_z$72–$L_z$73 kpc and $L_z$74 for $L_z$75–$L_z$76 kpc (Deng et al., 2024). This does not establish a unique origin for the warp, but it places GSE among the leading merger-driven explanations for a structure that has persisted for over $L_z$77 Gyr.

Taken together, these lines of evidence place GSE at the center of Milky Way formation studies. It is simultaneously a kinematic remnant, a chemically resolved disrupted dwarf galaxy, a globular-cluster system with episodic formation, a source of distant shells and retrograde streams, and a plausible driver of halo morphology, disk chemistry, and possibly the Galactic warp. The remaining controversies—especially progenitor mass, degree of internal stratification, and whether GSE is a single remnant or a composite structure—are now part of the subject rather than external caveats.