A long-running simulation finds a broader mix of masses among the first stars

The first stars helped determine how the young universe was illuminated and enriched with heavy elements. Their distribution of masses matters: massive stars produce intense radiation and can end their lives explosively, while sufficiently small stars could potentially survive to the present. Yet the paper notes that no confirmed metal-free stars have been identified in the Milky Way, leaving this distribution uncertain.

In a preprint posted to arXiv on November 5, Yasmine J. Meziani and colleagues investigate that question using a cosmological simulation. Their calculation produces a population less strongly weighted toward massive stars than several earlier simulations. It does not establish a universal mass distribution for the first stars; it follows one particular environment with important numerical and physical approximations.

Following successive generations of star formation

The team uses the FORGE’d in FIRE framework, running the GIZMO code to follow gravity, gas motions, magnetic fields, radiation transport and primordial chemistry. The model also includes feedback from individual stars through protostellar jets, winds, radiation and supernovae.

The simulated region is associated with the progenitor of an isolated ultra-faint dwarf galaxy. Starting with the first star formation near redshift 16, the calculation follows just over 500 million years of evolution, reaching approximately redshift 7. Stars form in bursts across five distinct dark-matter minihalos, rather than within one continuously active cloud.

Dense gas is refined to a mass resolution of about one-hundredth of the Sun’s mass and a spatial resolution of roughly one-thousandth of a parsec. Collapsing regions become numerical “sink particles” representing individual stars or protostars, which can accrete gas and generate feedback.

The analysis tracks 56 unique stars, including objects later combined through numerical mergers. Of these, 54 are strictly metal-free at the measurements used for the mass analysis. Two have trace enrichment and remain included in the population statistics, so the sample is not exclusively pristine.

A less extreme preference for giants

The central result concerns the initial mass function, or IMF: the relative numbers of stars at different masses. Above about 2.2 solar masses, the team fits a power law with a slope parameter of 1.9. In this convention, a smaller slope means a greater relative abundance of massive stars. The commonly used Kroupa IMF has a high-mass slope of 2.3, making the simulated population mildly “top-heavy” by that comparison.

This measurement uses each star’s mass when it first reaches the main sequence, or its maximum mass within the highest protostellar stage reached if it has not done so. Those are not necessarily final masses: some stars can continue growing after entering the main sequence.

The environments differ markedly. The first star-forming minihalo produces the most massive star, at approximately 110 solar masses. Its parent gas cloud is the most massive and has the highest surface density in the sample. Later minihalos produce maximum stellar masses of only about four to ten solar masses. The authors interpret this pattern as consistent with gas supply and feedback jointly limiting growth, although the small populations make comparisons uncertain.

Early snapshots also look more top-heavy than later ones. The mass distribution of objects present initially has slopes around 1.4–1.5, whereas a late snapshot reaches about 2.6. These snapshot distributions are distinct from the combined IMF of all stars formed. Their evolution suggests that comparing a short early calculation with a much longer history can produce different impressions of primordial star formation. The study does not isolate duration from other differences in physics and environment.

What remains unresolved

The simulation forms stars below one solar mass, including eight below an approximate 0.9-solar-mass survival threshold derived from a simple lifetime estimate. These are possible survivors, not observed ancient stars. Crucially, they lie below the estimated completeness limit of roughly one solar mass, where the calculation may miss additional objects and cannot reliably predict the population’s abundance or the IMF’s turnover.

To make the long calculation practical, the researchers artificially combine some close binaries. This is not a physical model of stellar collisions and can alter later dynamics, accretion and feedback. The model also omits a separate cosmological Lyman–Werner background, radiation that can destroy molecular hydrogen and delay cooling and collapse. Whether primordial stars launch the modeled protostellar jets remains uncertain, and stellar-remnant evolution is simplified.

The result is therefore a conditional picture: in this simulated environment, following multiple star-forming episodes yields a broader population than an early snapshot alone suggests. Establishing how widely that picture applies requires other environments and stronger checks on the unresolved physics.