A simulated primordial disk breaks into dozens of newborn stars

A collapsing cloud of primordial gas can produce a crowded stellar nursery rather than one isolated giant. That is the picture emerging from a new simulation by Vasilii Pustovoit and colleagues, which follows star formation inside an early ancestor of a Milky Way-like galaxy.

The study, posted to arXiv on November 5, 2026, and submitted to The Astrophysical Journal, is a preprint. Its findings describe a computer experiment, not an observation of the first stars. The calculation also has substantial limitations that affect how literally its stellar census should be read.

Following gas into a fragmenting disk

Population III stars are the first generation formed from primordial gas without heavier elements. Their masses and formation patterns depend on how that gas cools, collapses and divides into smaller structures.

The researchers used the FORGE’d in FIRE framework, combining large-scale galaxy physics with models of individual protostar formation. Their calculation follows gravity, gas dynamics, magnetic fields and radiation. Increasingly fine gas resolution lets them track dense collapsing material down to approximately astronomical-unit scales at the highest densities, although most disk gas is resolved more coarsely.

The selected halo contains about 20 million solar masses of dark matter at redshift roughly 21. In the parent cosmological simulation, it eventually becomes part of a Milky Way-mass system. The detailed calculation follows approximately 15,000 years after the first protostar forms.

Dense collapsing objects become “sink particles”: numerical stand-ins for protostars and their unresolved immediate surroundings. These particles can accrete gas, interact gravitationally and merge under specified conditions.

The initially collapsing structure develops spiral arms that break into new sinks. Two major fragmentation episodes occur around 3,500 and 8,000 years after the first protostar appears. Later formation sites lie farther out, as the inner region loses gas to growing protostars while the outer disk continues accumulating material.

Across the run, 62 sinks form and 30 disappear through mergers, leaving 32 at its end. Those numbers distinguish formation events from surviving objects; the simulation does not produce 62 enduring stars.

A growing population, not runaway individuals

By 15,000 years, the total mass in sinks reaches roughly 200 solar masses. Its growth follows approximately the square of elapsed time, consistent with theoretical expectations for accelerating accretion during self-gravitating turbulent collapse.

That acceleration belongs to the population as a whole. Individual sinks grow at roughly constant or declining rates, rather than showing runaway growth in which increasingly massive objects accrete ever faster. Local gas supplies and displacement from the disk limit their growth.

The most massive sink reaches about 25 solar masses. Lower-mass objects are more readily scattered away from the gas-rich region, after which their accretion largely stalls. Seven of the 32 survivors meet the study’s gravitational escape criterion at the final snapshot.

The surviving population has a “top-heavy” protostellar mass distribution: massive objects are relatively more common than in standard present-day stellar mass distributions. The fitted slope corresponds to approximately equal numbers per logarithmic mass interval. But this is an unfinished population, not a final initial mass function. None of the modeled objects has reached the main sequence.

The authors report a stellar mass fraction approaching 90% relative to stars plus remaining disk gas. That does not mean the halo has converted nearly all its gas into stars: the corresponding fraction is roughly half within a 0.1-parsec sphere and only about three parts in 100,000 across the whole halo.

Magnetic energy grows by several thousandfold, yet remains below the disk’s thermal energy. The field does not prevent vigorous fragmentation in this run. Without an otherwise identical nonmagnetic calculation, however, the researchers cannot isolate its effect or establish whether a turbulent magnetic dynamo is resolved.

Why the numbers remain provisional

Several modeling choices could alter the outcome. The reduced chemistry omits heating from molecular-hydrogen formation and cooling from its dissociation, as well as molecular line opacity. The modeled gas is colder than a reference primordial thermal track, potentially encouraging excess fragmentation. A Milky Way-like cosmic-ray background is also retained rather than adjusted to primordial conditions.

Protostellar jets encounter an implementation problem: their material mixes into nearby gas too readily, so they do not travel as far or disrupt infall as intended. The authors therefore regard stellar masses, accretion rates and efficiencies as overestimates. The adopted protostellar radii may also make early accretion heating too strong.

Finally, the parent simulation cannot resolve smaller halos that could have formed stars earlier. This is consequently a study of collapse in one relatively massive early halo, not a complete reconstruction of the region’s first star formation. Its strongest result is a detailed, conditional account of how spiral-arm fragmentation and a finite gas supply can build—and then reshape—a primordial protostellar population.