The First Stars Alone May Not Explain Tiny Galaxies’ Iron

A chemical puzzle in the smallest galaxies

The universe’s first stars formed from gas made almost entirely of hydrogen and helium. Their radiation, winds and explosions began spreading heavier elements into their surroundings. Ancient stars in ultra-faint galaxies can preserve traces of that enrichment, offering an indirect way to investigate a stellar population that is difficult to observe directly.

But there is a mismatch. Observations of nearby ultra-faint galaxies show iron abundances clustering around [Fe/H] = −2.5, with substantial scatter. That notation compares the iron-to-hydrogen ratio with the Sun’s; −2.5 means roughly one three-hundredth of the solar ratio. Many simulations predict considerably less iron. Whether the observed pattern represents a genuine plateau or increased scatter remains debated.

In an arXiv preprint submitted to The Astrophysical Journal, Pratik J. Gandhi and colleagues test whether adding Population III—the first generation of stars—can resolve the discrepancy. Their answer is conditional: not reliably under the prescriptions they explored.

Modelling the first stellar generations

The team used FIRE-3 cosmological zoom-in simulations, which follow selected regions in high detail while retaining their wider cosmic setting. They studied an isolated ultra-faint galaxy and populations around a system with a halo mass comparable to the Large Magellanic Cloud’s.

Their Population III prescription approximates a population containing only massive stars, between eight and 100 times the Sun’s mass, at sufficiently low heavy-element abundance. A smooth transition connects this prescription to the ordinary stellar population model as enrichment increases. These massive-star populations produce stronger radiation, winds and core-collapse supernova feedback per unit stellar mass.

This is a simplified numerical prescription, not a direct calculation of individual primordial stars. Most variants retain the standard FIRE-3 supernova yields rather than using element-production patterns unique to Population III stars.

The finest gas mass resolutions were about 30 solar masses in the isolated simulation and 110 solar masses in the larger environment. Most runs reached the present day; the highest-resolution larger-environment run stopped at redshift four, after its ultra-faints had largely ceased forming stars.

An enrichment floor is not enough

The Population III models introduce a minimum heavy-element abundance among surviving stars. Depending on the assumed transition threshold, this effective floor lies around one hundred-thousandth to one thousandth of the Sun’s total heavy-element abundance. The corresponding iron floors are approximately [Fe/H] = −7 to −5.

Importantly, this floor partly follows from the model’s construction: its massive Population III stellar populations do not survive to the present day. It is therefore not an observed universal minimum, nor an independent prediction unaffected by the assumed stellar population.

Raising the transition threshold raises the floor, but does not consistently bring the simulated galaxies into agreement with observations. In four otherwise identical reruns of the baseline model, only about 40 percent of ultra-faints reached [Fe/H] above roughly −3.5. Individual galaxies differed between runs, while the broad under-enrichment problem remained.

Later explosions change the outcome

The more iron-rich simulated ultra-faints generally formed successive generations of stars over longer periods after Population III enrichment. That allowed Type Ia supernovae—explosions involving white dwarfs—to contribute iron that subsequent stars incorporated.

Galaxies with brief star-formation histories instead remained dominated by early core-collapse enrichment. They could accumulate substantial amounts of other heavy elements without reaching comparable iron abundances. Iron is therefore an incomplete proxy for their overall chemical enrichment.

Among the tested prescriptions, combining Population III feedback with an increased Type Ia supernova rate at low metallicity gave the best overall agreement with both typical iron abundances and their spread within galaxies. An explicitly exploratory model boosting Population III supernova iron yields tenfold also improved agreement. Neither result establishes that these adjustments describe nature uniquely.

Magnesium offers another test

The simulations predict a different pattern for magnesium, which comes predominantly from core-collapse supernovae. Across the model variants, ultra-faint galaxies reached at least approximately [Mg/H] = −4, even when their iron abundance was much lower. Magnesium depended less strongly on how long star formation continued. This predicted stellar mass–magnesium relation offers an additional observational test.

Several limitations remain. The simulations use a spatially uniform reionization background, omit an explicit Lyman–Werner radiation background that can disrupt molecular hydrogen, and lack a Milky Way– or Local Group–like environment. Some variants also produce internal iron spreads larger than observed. The study does not establish why particular galaxies sustain star formation longer.

The central lesson is therefore model-specific but consequential: understanding the first stars’ chemical legacy also requires understanding the later generations that inherited and altered it.