Scientists Identify Stellar Remnants of Galaxy the Milky Way Consumed Billions of Years Ago

The Milky Way is not a monolithic structure. It is, in part, a composite of smaller galaxies consumed over billions of years—a fact that astronomers are only now able to piece together in detail. The latest evidence: a group of 20 ancient stars, chemically distinct from their neighbors, sitting in an unexpected location within our galaxy's disk. Researchers believe they are the scattered bones of a dwarf galaxy that was devoured approximately 10 billion years ago [1][2].

The team, led by Federico Sestito, has named the lost galaxy "Loki," after the Norse trickster god. "Our accreted stars gave us some hard time in understanding their origin," Sestito explained, drawing the parallel to the mythological figure whose intentions are notoriously difficult to decipher [3].

The Discovery: 20 Stars Where They Shouldn't Be

The Loki candidates were identified through a combination of astrometric data from the European Space Agency's Gaia spacecraft (Data Release 3) and high-resolution spectroscopy from the ESPaDOnS instrument on the Canada-France-Hawaii Telescope (CFHT), located near the summit of Maunakea, Hawaii [2][3].

What makes these 20 stars unusual is twofold: their chemistry and their location. Most extremely metal-poor stars—those containing very low abundances of elements heavier than hydrogen and helium—reside in the galactic halo, the diffuse spherical region surrounding the disk. Metal-poor stars are among the oldest objects in the universe because heavy elements are produced by successive generations of stellar explosions. Finding a cohort of them embedded within the disk itself, roughly 7,000 light-years from our solar system, challenges standard models of where ancient merger debris should end up [2][4].

The team measured 23 chemical elements in each star, including carbon, magnesium, calcium, titanium, strontium, barium, and europium. Despite differences in their orbital trajectories—11 stars move in the same direction as most disk stars (prograde), while 9 move in the opposite direction (retrograde)—all 20 share remarkably similar chemical abundance patterns. The dispersion in their element-to-iron ratios ([X/Fe]) is narrower than what is seen in typical halo or bulge populations at the same metallicity [3][4][5].

This chemical homogeneity is the strongest argument for a common origin. The enrichment pattern shows signatures of high-energy supernovae and hypernovae, contributions from fast-rotating massive stars, and neutron star mergers—but notably lacks the signature of Type Ia supernovae (white dwarf explosions). This absence suggests Loki's star-forming life was cut short before slower evolutionary channels could contribute [4][5].

How Big Was Loki?

Models used by the research team estimate Loki's total mass at approximately 1.4 billion solar masses [1][3]. For perspective, this places Loki in a specific weight class among known devoured galaxies:

Source: Kruijssen et al. 2020, MNRAS; Sestito et al. 2026, MNRAS

Data as of May 23, 2026CSV

The Large Magellanic Cloud (LMC), the Milky Way's largest surviving satellite galaxy, has a total mass of about 138 billion solar masses including dark matter, and a stellar mass of roughly 2-3 billion solar masses [6]. Loki, at 1.4 billion solar masses total, was considerably smaller—perhaps one-tenth the LMC's stellar mass and a fraction of its total mass. It was larger, however, than most of the other identified merger remnants except Kraken, the earliest known accretion event, estimated at about 2 billion solar masses in stellar mass [7].

The Gaia-Enceladus/Sausage (GES) event—until now the most discussed major merger—involved a galaxy with a stellar mass of roughly 300 million to 1 billion solar masses, depending on the study [8][9]. If the 1.4 billion solar mass estimate for Loki holds, it was comparable to or somewhat more massive than GES in total mass, making this a significant addition to the Milky Way's formation history.

Loki vs. Gaia-Enceladus: A Distinct Event

A central question is whether Loki represents a genuinely new discovery or merely additional debris from the already-documented Gaia-Enceladus/Sausage merger, which was first identified in 2018 [10].

The evidence points toward a distinct event. GES remnant stars are characterized by highly eccentric orbits (eccentricity greater than 0.7), low angular momentum, and radial velocities reaching 400 km/s. They populate the inner stellar halo and contributed an estimated 41-74% of the stellar population within 30 kiloparsecs of the galactic center [10][11]. Loki's stars, by contrast, are embedded in the galactic plane itself, on stretched-out but relatively flat orbits. Their chemical fingerprint—particularly the absence of Type Ia supernova contributions—also differs from GES stars, which show evidence of more prolonged chemical evolution [4][5].

The study was published in Monthly Notices of the Royal Astronomical Society (DOI: 10.1093/mnras/stag563) in early 2026, placing it within the peer-review process of one of astronomy's premier journals [5]. However, independent confirmation from other research groups using different datasets has not yet appeared—a limitation the authors acknowledge.

The Milky Way's Merger Catalog

Loki joins a growing list of identified accretion events in the Milky Way's history. Using the ages, metallicities, and orbital properties of globular clusters, researchers have reconstructed a rough chronological sequence of major mergers [7]:

Source: Kruijssen et al. 2020; Sestito et al. 2026

Data as of May 23, 2026CSV

The Milky Way cannibalized at least five significant dwarf galaxies: Kraken (~11 billion years ago), the Helmi Streams progenitor (~10 billion years ago), Sequoia (~9.3 billion years ago), Gaia-Enceladus (~9 billion years ago), and Sagittarius (~7 billion years ago, and still ongoing) [7][12]. Loki would slot in at roughly 10 billion years ago, contemporary with or slightly before Gaia-Enceladus.

How Many Stars Are Immigrants?

The question of what fraction of the Milky Way's roughly 200-400 billion stars originated outside the galaxy is actively debated. Studies combining Gaia astrometry with spectroscopic surveys like APOGEE and the H3 Survey have found that accreted stars dominate the stellar halo. The H3 Survey, which gathered 6D phase-space and chemical information for 5,684 giant stars within 50 kiloparsecs, found that the stellar halo is "entirely comprised of substructure"—effectively all of it appears to have been accreted [13].

However, the halo represents a small fraction of the Milky Way's total stellar mass. Simulations of Milky Way-mass galaxies suggest that accreted stars constitute roughly 10% of total stellar mass, with the vast majority of stars having formed in situ [14]. The total stellar mass accreted from known events is estimated at approximately 2.2 billion solar masses [7]. If Loki contributes an additional 1.4 billion solar masses, the total from identified mergers rises to roughly 3.6 billion solar masses—still under 5% of the galaxy's estimated total stellar mass of 50-70 billion solar masses.

This does not fundamentally overturn the picture of the Milky Way as a galaxy that grew primarily through internal star formation. But it does complicate the narrative: the thick disk, the halo, and even portions of the thin disk carry the fingerprints of past collisions. The Milky Way is less a single organism and more a composite body that absorbed its neighbors.

Why Now? The Technology That Made This Possible

Galactic archaeology as a field dates back decades, but the tools available today are qualitatively different from those of the 1990s and 2000s. Three developments converged to make discoveries like Loki possible:

Gaia: ESA's Gaia spacecraft, launched in 2013, has provided precise positions, distances, and proper motions for nearly two billion stars across successive data releases. Gaia DR3, used in the Loki study, includes radial velocities for 33 million stars and photometric data sufficient for Bayesian distance estimates to metal-poor targets [2][15].

High-resolution spectroscopy at scale: Instruments like ESPaDOnS on CFHT can measure dozens of chemical elements per star, but they observe one star at a time. Upcoming multi-object spectrographs like WEAVE (on the William Herschel Telescope) and 4MOST (on ESO's VISTA telescope) will observe thousands of stars simultaneously, dramatically expanding the catalog of chemically characterized metal-poor stars [5].

Computational advances: Machine learning frameworks like BINGO now determine stellar ages from chemical abundances, and N-body simulations can reconstruct plausible merger histories from the present-day distribution of globular clusters and field stars [7][11].

Previous generations of astronomers lacked the combination of all-sky astrometry, precise radial velocities, and detailed chemical abundances for large samples. Without all three dimensions—spatial, kinematic, and chemical—disentangling overlapping populations is extremely difficult.

Source: OpenAlex

Data as of Jan 1, 2026CSV

The explosion in galactic archaeology publications since 2018 (peaking at 83 papers in 2023) directly correlates with Gaia data releases. Before Gaia, identifying streams and substructure required painstaking follow-up of individual candidate stars. Now researchers can systematically survey populations and let chemistry and kinematics reveal common origins.

The Skeptical View

Not all astronomers are fully convinced that the 20 stars represent a single accreted system. The primary uncertainty lies in distinguishing genuine accretion debris from rare native Milky Way stars that happen to share unusual properties.

Metal-poor stars in the disk could, in principle, have formed in the early Milky Way itself and been scattered onto unusual orbits by subsequent perturbations. The absence of Type Ia supernova signatures is consistent with an accreted origin—but it is also consistent with any stellar population that formed in a short burst, regardless of environment [4].

Spectroscopic dating of individual stars carries uncertainties of 1-2 billion years at these ages, making it difficult to establish whether all 20 stars are truly coeval [3]. The sample size of 20, while sufficient to establish chemical homogeneity, is small enough that stochastic effects in nucleosynthetic enrichment cannot be entirely ruled out.

The authors note that future large surveys (WEAVE, 4MOST) will be critical for testing the Loki hypothesis by identifying additional candidate members and establishing whether the chemical signature is genuinely unique or shared with other populations [5].

What Past Mergers Tell Us About the Andromeda Collision

The Milky Way and Andromeda galaxy (M31) are approaching each other and will interact in approximately 4.5 billion years. Recent simulations have revised the probability of a direct merger to roughly 50%, down from earlier estimates of near-certainty, due to updated measurements of Andromeda's tangential velocity [16][17].

Past merger dynamics offer partial—but limited—predictive power. The Gaia-Enceladus collision involved a mass ratio of roughly 4:1 (Milky Way to GES), producing dramatic heating of the existing disk and contributing to the formation of the thick disk [11]. A Milky Way-Andromeda interaction would involve a mass ratio closer to 1:1, making it a qualitatively different event.

What past mergers do reveal is the timescale and aftermath: the GES merger, for instance, took 1-2 billion years from first interaction to full disruption. The resulting starburst from compressed gas, followed by redistribution of stars onto new orbits, provides a template for what happens to stellar populations during galactic collisions. The supermassive black holes of the two galaxies would eventually coalesce within about 16 million years of the final merger [18].

However, extrapolating from dwarf galaxy accretions (mass ratios of 10:1 to 100:1) to a major merger (1:1) has significant limitations. The gravitational dynamics, tidal forces, and timescales all scale non-linearly with mass ratio. Past mergers are instructive analogs, not direct previews.

The Bigger Picture

The identification of Loki reinforces a broader conclusion about galaxy formation: the hierarchical model, in which large galaxies are built up from smaller ones over cosmic time, continues to find observational support. Each new discovery adds resolution to our understanding of the Milky Way's assembly history.

Twenty stars may seem like a modest haul, but they represent the identifiable tip of what was once an entire galaxy. The chemical and kinematic information encoded in those stars—formed 10-12 billion years ago, during the first 2-3 billion years after the Big Bang—preserves a record of conditions in the early universe that no longer exists in any intact form [2][4].

Whether Loki survives independent scrutiny or is eventually reclassified as part of a known structure, the methodology and tools that enabled its identification will continue to reshape our understanding of how the Milky Way—and galaxies in general—came to be.