Simulations map the changing gas supply of the first galaxies

Understanding early galaxies requires more than counting their stars. Much of their ordinary matter—the material astronomers call baryons—remains gaseous, moving between different temperatures and chemical states. Gas falls into galaxies, cools and forms stars; stellar radiation, explosions and winds then heat or expel some of that material.

Umberto Maio and Céline Péroux explore this cycle using the ColdSIM simulations. Their study, listed in the arXiv record as published in Astronomy & Astrophysics, presents theoretical predictions rather than a direct observational inventory of all early gas.

Following chemistry as galaxies grow

ColdSIM combines gravity and gas dynamics with time-dependent chemistry. Rather than estimating molecular hydrogen only after a simulation has finished, it follows the molecule’s formation and destruction as cosmic structures evolve. The calculations also include cooling, heating, star formation, stellar lifetimes, chemical enrichment and stellar feedback—the energy and material that stars release into their surroundings.

The researchers analyse a reference simulation and compare it with higher-resolution and larger-volume runs. They distinguish material inside haloes’ virial radii from material outside those boundaries. Their inventory of bound material therefore includes both galaxies and their surrounding circumgalactic gas, not just their central star-forming regions.

They classify gas as cold below 10,000 kelvin, warm between 10,000 and 10 million kelvin, and hot above 10 million kelvin. These are broad modelling categories: “cold” does not necessarily mean the very low temperatures of a dense molecular cloud.

A changing balance of gas

At redshifts above about six, cold gas dominates the simulated cosmic baryon budget. Near the completion of reionisation—the process in which radiation strips electrons from largely neutral intergalactic gas—the balance changes. Stellar feedback and increasing ultraviolet radiation heat the gas, and the warm component becomes the largest reservoir at later times.

The distinction between the cosmic inventory and material inside haloes matters. Most ordinary matter in these simulations lies outside bound structures, so adding up galaxies and their immediate surroundings does not recover the full cosmic budget. Within bound structures, cold and warm gas account for approximately 30% and 70% of the total bound gas mass, respectively, by redshift five. Hot gas remains a minor component.

Neutral atomic hydrogen, known as HI, closely tracks the cold phase. Molecular hydrogen, H2, responds more strongly to local conditions: it forms in cooling gas but can be destroyed by radiation and feedback. Measurements of atomic and molecular gas therefore need not reveal a galaxy’s entire gas supply, particularly when substantial warm gas is present.

Young stars have returned less material

Another result concerns the stellar return fraction: the share of a population’s initial stellar mass released back into surrounding gas through stellar evolution.

The simulations predict values around 15–20% at redshift five, roughly half the commonly adopted estimates discussed in the paper. The explanation is time. Young populations have not yet passed through all the evolutionary stages that return material to space. The predicted fraction is smaller still at earlier epochs.

This calculation assumes a Salpeter initial mass function, which specifies the relative numbers of stars born at different masses. Other assumptions about stellar populations affect inferred masses and recycling. The result supports accounting for stellar ages rather than adopting one universal replacement value.

Returned stellar material is not, however, the main source of the mass changing from cold to warm gas. The authors emphasise the energy injected by feedback: a relatively small amount of star formation can heat a much larger surrounding reservoir.

Short depletion times do not mean complete conversion

The team calculates gas depletion times by dividing a gas reservoir’s mass by its current star-formation rate. Molecular depletion times can reach 10–100 million years, but the distributions are broad and vary with epoch and galaxy properties. The paper reports that the bulk of molecular depletion times lie around 10–50 million years at redshift 16 and 100 million to one billion years at redshift six. Atomic-gas depletion times are generally longer.

These ratios describe consumption at the current rate, not a prediction that an evolving galaxy will actually run out of gas on schedule. Accretion, heating, cooling and changing star formation can all alter the outcome.

Nor do short depletion times mean that all available gas rapidly becomes stars. The study’s median star-formation efficiency, defined as gas free-fall time divided by depletion time, remains around 1–4% for the overall population. The more actively star-forming subset reaches values around 10%.

The predictions depend on feedback prescriptions, stellar-population assumptions, numerical resolution and the adopted ultraviolet background and reionisation history. Gas relations also show substantial scatter. Comparisons with surveys at later cosmic epochs provide context, but do not directly verify the earliest predictions. The study offers a framework for interpreting difficult observations—not a completed observational census of the young universe.

Publication and source record

The original arXiv Atom record and submission history establish the first submission as 27 February 2026 at 19:00:00 UTC, corresponding to the supplied published_at value of 2026-02-27T19:00:00.000Z. Revision v4 was submitted on 6 November 2026 at 16:35:24 UTC; neither timestamp should be confused with the journal’s publication date.

The arXiv record also supplies the publication comment, journal reference A&A 708, A244 (2026), and journal DOI. Its abstract matches the supplied metadata abstract verbatim. That historical arXiv abstract is distinct from the PDF’s structured abstract and should remain unchanged in source metadata; this article’s summary is an editorial paraphrase.