Nearby Dwarf Iron Is a Fossil of the First Stars

Four MEGATRON papers published on 30 September 2026 show that pair-instability blasts from the first stars can set the iron content of the tiniest nearby galaxies. The suite, co-led by Martin Rey at the University of Bath, Harley Katz at the University of Chicago, and Corentin Cadiou at the Institut d’Astrophysique de Paris, follows a Milky Way-mass progenitor from metal-free gas to about redshift 8.

The point is not only a sharper view of cosmic dawn. It is that ultra-faint dwarfs around the Milky Way, and the metal-poor stars in its halo, can now be read as the same physics the James Webb Space Telescope records in infant galaxies, rather than as a separate local puzzle.

The Iron Floor in the Tiniest Dwarfs

Observed ultra-faint dwarfs with stellar masses of 100,000 Suns and below sit on a flat iron sequence near [Fe/H] of about -2.5, a few hundred times less iron than the Sun. That floor had been a problem: the galaxies are so small that later star formation should not have left them this iron-rich, and several explanations have been on the table, from a changed later-star mass mix to iron blown in from outside.

Rey’s team tracked more than 500 dwarf galaxies in MEGATRON, a cosmological radiation-hydrodynamics suite run with the RAMSES-RTZ code and the PRISM interstellar-medium model. The simulated dwarfs match the local stellar mass-iron relation, including an over-abundance along that same floor, which the paper ties to chemical enrichment by Population III pair-instability supernovae.

THE IRON SEQUENCE MEGATRON RECOVERS

Figure 3 of the paper finds that 78 percent of the simulated faint dwarfs cluster on that iron sequence. Both the floor and the iron-poor tail still appear when the team changes its assumptions for later, metal-poor (Population II) feedback, and both survive among bound satellites around the central galaxy at redshift 0, which is the comparison that matters for the real Milky Way.

Almost Every Faint Dwarf Hosts One First-Star Blast

The chemistry is sparse because the first stars are sparse. In the MEGATRON dwarfs, later core-collapse and hypernova blasts from Population III objects are rare, a consequence of the top-heavy first-star mass function the team adopted. What remains is a handful of very massive explosions.

HOW THE FIRST BLASTS SET THE FLOOR

• One explosion: Almost every faint dwarf undergoes exactly one Population III blast.
• External pollution: About 2 percent of the dwarfs have no Population III explosion and are enriched only by diluted ejecta from neighbours.
• High-mass pair-instability: Dwarfs on the plateau almost exclusively host explosions of stars above 160 and up to 300 solar masses.
• Lower-mass pair-instability: A separate bin covers progenitors from 140 up to 160 solar masses, which do not dominate the plateau.

That is a sharp prediction for stellar archaeologists. If the smallest Milky Way satellites really are one-blast fossils, their iron, and the detailed mix of other metals in their oldest stars, should look like the yield of a single very massive first star, not like a long series of ordinary supernovae.

A Strong Ultraviolet Fog From the Protogalaxy

Pair-instability ejecta only help if the host halo is deep enough to keep them. MEGATRON’s iron plateau in the smallest dwarfs appears because a strong Lyman-Werner background from the protogalaxy pushes those blasts into haloes of about 10 million solar masses, massive enough to retain the iron.

Lyman-Werner photons break molecular hydrogen, the coolant that lets tiny minihaloes form stars. Among the three processes that shut off first-star formation in those minihaloes (the ultraviolet fog, gas starvation, and metals arriving from outside) the Lyman-Werner field is the strongest, according to a related MEGATRON study led by Anatole Storck at the University of Oxford.

The first Population III stars still form in minihaloes of about 100,000 to 1 million solar masses near redshift 30, where cooling is dominated by molecular hydrogen. Once the fog is up, most later first stars form in more massive, atomic-cooling systems. The global first-star formation rate then settles at 0.001 solar masses per year by redshift 20.

A few haloes get a second chance if their earlier stars collapse straight to black holes and leave the gas metal-free. If a halo grows enough, it can form up to about 100 Population III stars in one burst, a target Storck’s team says JWST could see with moderate gravitational lensing. Only 0.06 percent of the simulated first stars form inside the virial radius of galaxies brighter than ultraviolet magnitude -17, so the typical birth site is not the core of a JWST-bright galaxy.

Where the First Stars End Up Today

The same zoom region is a Milky Way analogue, which is why the local fossil record is in play. Dark-matter particle mass is 25,000 solar masses and the gas is resolved at about 3 parsecs, fine enough to capture most first-star birth sites in the smallest galaxies. The initial conditions, taken from the VINTERGATAN-GM suite, grow a halo of about 1 trillion solar masses by redshift 0, with a stellar mass of about 50 billion solar masses if the same region is evolved with tested low-redshift models.

FROM COSMIC DAWN TO THE GALACTIC HALO

1. Around redshift 30: First stars form in molecular-hydrogen minihaloes of about 100,000 to 1 million solar masses.
2. Redshift 20: The global Population III star-formation rate settles at 0.001 solar masses per year, with Lyman-Werner radiation the main throttle in minihaloes.
3. About redshift 8: The published high-redshift runs stop, after tracking gas, starlight, and metals in a region destined to become a Milky Way-mass system.
4. Redshift 0: First-star remnants are traced in a matching dark-matter-only run; most sit in the analogue’s stellar halo.

Storck’s team finds that 75 to 80 percent in the stellar halo of the simulated Milky Way analogue, with the rest bound to lower-mass systems, including satellite haloes. Upcoming halo surveys such as 4MOST and WEAVE are named in that paper as the practical hunt: if those remnants carry a pair-instability fingerprint, it should show up in metal-poor halo stars and in the faintest satellites, not only in JWST spectra of galaxies 13 billion years in the past.

No surviving metal-free star has been found in the Milky Way or its dwarfs, which already sets a lower bound on the first-star mass function near 0.7 to 0.8 solar masses. MEGATRON does not overturn that non-detection. It says where the dead ones, and the chemical ash of the very massive ones, should be sitting.

A Library of 175,000 Galaxy Spectra

The other half of the bridge is what JWST actually measures: emission lines and continua from gas in galaxies at redshift 6 and beyond. Katz’s introductory MEGATRON paper presents a library of more than 175,000 simulated galaxy spectra and argues that much of the diversity JWST already sees arises naturally in a standard cosmology, once radiation, metals, and molecules are evolved together rather than painted on later.

The team describes the suite as the first cosmological runs to couple a large non-equilibrium network of primordial species, metals, and molecules to multifrequency radiation transport on the fly, starting from zero metals and resolving haloes well below the atomic-cooling limit. Simplified models, the Bath release says, can underestimate how starlight and chemical change shape the gas around galaxies, and miss structures that appear once those effects are resolved.

The James Webb Space Telescope gives us a direct glimpse of the infant cosmos, while stellar archaeology allows us to study the relics of those earliest times in our own Galactic neighbourhood. MEGATRON provides a physical bridge between the two.

Martin Rey, Department of Physics, University of Bath

That bridge has to cross a mismatch in elements. JWST spectra of early galaxies are rich in oxygen lines from hot gas; local metal-poor stars are graded on iron. MEGATRON carries both, so a dwarf’s iron floor and a high-redshift galaxy’s oxygen lines can be asked whether they come from the same enrichment path.

Two of the other 30 September papers show why the extra physics is not cosmetic. Cadiou’s circumgalactic study compares the full non-equilibrium calculation plus local radiation with the usual post-processed equilibrium models that assume a uniform ultraviolet background. Ions shift, and hydrogen damped Lyman-alpha covering fractions that differ by 40 percent appear once recombination lags and local radiation anisotropy are allowed. Katz’s ultraviolet-slope paper finds galaxies with spectral slope beta steeper than -2.8 that are not all leaking ionizing photons: a second population has an escape fraction below 1 percent, because low gas densities delay nebular emission past the lives of the massive stars.

THE FOUR OPEN JOURNAL OF ASTROPHYSICS PAPERS

The papers are explicit about what is missing. The runs include no active galactic nuclei, and the adopted stellar-population and chemical-yield models are limited. Those caveats sit next to the claim, they do not cancel it: the diversity of JWST spectra is being used as a test of the gas physics, not as a finished census of every early galaxy.

Forty Million Hours and the Physics Still Missing

MEGATRON started in 2023 and is scheduled to run until 2030. The 30 September set is the collaboration’s first large published block, with more papers expected. The high-redshift suite uses four simulations of the same initial conditions, each with a different Population II star-formation and feedback model and the same Population III treatment, which is how the iron floor can be called stable against later-star assumptions.

WHAT THE PUBLISHED SUITE ACTUALLY RESOLVES

• Chemistry network: At least 80 atoms, ions, and molecules in the galaxy-formation model; the circumgalactic paper specifies 81 ions and molecules.
• Grid scale: About 3 parsecs in the high-redshift zoom, with circumgalactic cold and warm gas reaching 20 parsecs and averaging 200 parsecs at cosmic noon.
• Volume: A 20 cubic comoving megaparsec per h zoom inside a 50 comoving megaparsec per h box, chosen as an early-forming Milky Way progenitor.
• Next award: Rey’s group at Bath has 40 million processor hours on UK supercomputers for higher-resolution follow-on runs, described as equal to five million laptops in parallel for a year.

Rey said the elements that make the world and life possible, carbon, oxygen, iron and many others, were forged by stars, and that MEGATRON is meant to test how the first stars formed and enriched their surroundings against JWST and against the chemical fingerprints in ancient stars. In the Bath statement he also said the two datasets now let researchers test competing models of the first stars in ways that were not possible before.

The competing models are no longer abstract. If the faintest Milky Way satellites are one-blast pair-instability fossils, their iron should stay near that floor, a minority should be still more iron-poor, and most of the first-star ash in a Milky Way-mass system should already be in the stellar halo, waiting on surveys that can grade those stars one by one.