For years, computer simulations of the earliest galaxies have struggled to match what telescopes actually show. Depending on the model and the epoch in question, simulations sometimes produced too few bright galaxies, or galaxies that were too small, too quiescent, or too gas-poor compared with observations. Some models overproduced bright galaxies at certain redshifts while underproducing them at others. Now a team of astrophysicists believes it knows why the models keep missing the mark: two physical processes that most simulations left out.
The COLIBRE project, led by Joop Schaye and Evgenii Chaikin, has produced a new generation of cosmological simulations that explicitly track cold gas in the interstellar medium and the full life cycle of cosmic dust. Their results, accepted for publication in Monthly Notices of the Royal Astronomical Society in spring 2026, argue that these ingredients reshape how virtual galaxies assemble, bringing simulations significantly closer to observed reality. The work was highlighted by the Royal Astronomical Society in May 2026.
What earlier simulations got wrong
The core issue is a computational shortcut that dates back roughly two decades, though some research groups had already begun relaxing these constraints before COLIBRE. To keep simulations tractable, modelers imposed artificial temperature and pressure floors on gas, preventing it from cooling below a few thousand degrees Kelvin. The reason was practical: tracking gas at lower temperatures demands far more computing power and, without careful handling, can produce unphysical results. But the COLIBRE team argues in their central paper that this shortcut masked real physics that governs how galaxies form.
Two processes stand out. The first involves cold, multiphase gas in the interstellar medium. Dense, cold pockets of gas are where stars actually form. When simulations smooth this gas into a warm, uniform medium, they miss the conditions that control star-formation rates. In reality, pressure support weakens in the densest regions, gravity wins more easily, and star-forming clouds collapse in patterns that a temperature floor simply cannot reproduce.
The second is dust. A companion paper details an explicit dust life-cycle model coupled to cold-gas chemistry. Dust grains are not passive byproducts of stellar evolution. They shield gas from ultraviolet radiation, promote molecule formation on their surfaces, and alter cooling rates. They grow in dense, metal-rich regions by accreting material from surrounding gas, and they are destroyed by shocks, sputtering, and intense radiation near young stars and supernovae. When simulations track these effects directly, the resulting galaxies develop more realistic distributions of cold gas, metals, and molecular clouds.
A third ingredient: feedback before the explosion
A separate study within the COLIBRE framework examines a process that most earlier models also underrepresented: pre-supernova feedback. Before massive stars explode, they inject energy into surrounding gas through ionizing radiation and thermal pressure. This slower buildup of heat from HII regions precedes the first supernovae by millions of years.
Tests on isolated disk galaxies at multiple resolutions showed that this continuous energy injection regulates star formation in ways that post-explosion feedback alone cannot replicate. Including early feedback keeps gas more diffuse, limits runaway starbursts, and produces disk galaxies that better resemble the spirals and irregulars astronomers observe. Most prior models treated stellar feedback as a single explosive event, collapsing a drawn-out physical process into an instantaneous punch.
The calibration strategy behind COLIBRE has also been published, in MNRAS volume 548. Rather than hand-tuning feedback parameters until outputs looked right, the team used an emulator-based search to systematically explore which combinations of feedback strength reproduce measured galaxy properties. This approach lets them identify models that simultaneously fit galaxy stellar masses, sizes, and star-formation histories within a single consistent framework.
Where the results converge
Across these studies, the findings point in a consistent direction. When cold gas, dust physics, and pre-supernova feedback are modeled explicitly, simulated galaxies form stars more gradually, retain more cold material in their disks, and develop internal structures that more closely match observations. These processes also interact with each other: dust-enhanced cooling promotes cold-gas formation, while early stellar feedback prevents that gas from collapsing too quickly. Removing the artificial temperature floor allows these competing effects to play out naturally, rather than being overridden by a numerical approximation.
Open questions and limits
The COLIBRE results have passed peer review, but several uncertainties remain. The project has not yet released raw simulation data publicly, so independent groups cannot run their own analyses on the outputs. The team’s claims about improved agreement with observations are based on internal comparisons described in the papers, not on head-to-head tests against specific datasets from the James Webb Space Telescope or other recent high-redshift surveys. Whether the improvements hold up against the flood of new JWST galaxy data is a test the broader community will need to perform.
There is also a question of generalizability. The pre-supernova feedback tests were conducted on isolated disk galaxies, a controlled setting that strips away the chaotic environment of a full cosmological volume. The team’s cosmological runs include these physics, but the controlled tests that demonstrate the effect most cleanly are limited in scope. Galaxy interactions, mergers, and dense cluster environments may alter how efficiently early feedback couples to gas. It is not yet clear whether a single feedback prescription can capture behavior across dwarf galaxies, Milky Way-scale systems, and the most massive ellipticals.
Dust modeling introduces its own ambiguity. The dust life-cycle paper establishes the physical mechanisms, but translating those effects into specific predictions about galaxy luminosity functions or spectral properties at high redshift requires additional steps. Radiative transfer calculations, which convert distributions of stars, gas, and dust into synthetic images and spectra, carry their own uncertainties. Until those calculations are carried out systematically and compared with observations, it will be difficult to say precisely how much of the mismatch between older simulations and JWST data traces to dust physics alone.
Resolution remains a constraint as well. Even with modern supercomputers, cosmological simulations cannot resolve individual stars or small molecular clouds. They rely on “subgrid” recipes to describe how unresolved gas turns into stars and how feedback couples to surrounding material. COLIBRE improves the physical basis of those recipes, but they remain approximations. The team reports resolution tests indicating that key results are stable over a reasonable range, though some quantities, such as the detailed structure of the cold interstellar medium or the properties of the smallest dwarf galaxies, may shift as future runs push to higher resolution.
It is also worth noting that COLIBRE is not the only simulation project pursuing more realistic interstellar-medium physics. Projects like FIRE and SIMBA have incorporated elements of cold-gas and feedback modeling, though with different numerical approaches. COLIBRE’s contribution is to combine cold-ISM tracking, a full dust life cycle, and pre-supernova feedback within a single large-scale cosmological framework, a combination that had not been attempted at this scale before.
What this means for the race to explain early galaxies
The practical significance of this work is tied to timing. Telescopes, JWST chief among them, are now delivering unprecedented views of galaxies that formed within the universe’s first few hundred million years. Theorists are racing to update their models to keep pace with what those instruments reveal. COLIBRE offers one of the first large-scale attempts to fold a more realistic interstellar medium, detailed dust evolution, and early stellar feedback into fully cosmological simulations.
The Royal Astronomical Society’s summary confirms the core narrative from the research team: cold gas and dust were missing from earlier simulations, and adding them changes the results. But press releases by nature emphasize positive findings and omit caveats present in the technical papers. The lay explanation that prior simulations imposed temperature floors due to computational complexity, for instance, is accurate but does not convey that those floors were understood as deliberate approximations, not oversights. Many earlier models were tuned to match observations despite those simplifications, which complicates any clean before-and-after comparison.
The COLIBRE results provide a concrete, testable claim: that previously neglected small-scale physics can substantially alter the large-scale appearance of galaxies across cosmic time. The results do not settle every debate about galaxy formation, and they leave room for alternative explanations of remaining discrepancies. But as more groups compare these simulations to the latest telescope data, the field will be able to judge how much of the universe’s early complexity can be explained by cold gas and dust, and how much still demands entirely new ideas.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
