For decades, computer models of galaxy formation have relied on a significant shortcut: preventing the simulated gas inside galaxies from cooling below roughly 10,000 degrees Fahrenheit, hotter than the surface of the Sun. The workaround was necessary because modeling colder gas requires tracking far more complex physics and chemistry. Yet cold gas is precisely where stars are born, and its absence has long been a recognized limitation of even the most sophisticated models.
COLIBRE eliminates that compromise. The simulations, developed over nearly a decade by an international team spanning Europe, Australia, and the United States, allow gas to cool to as low as 10 Kelvin, close to absolute zero, directly modeling the cold interstellar medium that drives star formation.
Cold Gas and Dust Finally Enter the Picture
The key technical advance in COLIBRE is its simultaneous treatment of cold gas and the tiny solid particles, dust grains, suspended within galaxies. These grains play an outsized role in galactic chemistry: they provide surfaces on which hydrogen molecules form, shield gas from ultraviolet radiation that would otherwise prevent cooling, and determine how galaxies appear to telescopes by absorbing starlight and re-emitting it in the infrared.
Five nested simulation volumes spanning 25 to 400 comoving megaparsecs, colour-coded by gas and stellar surface density. ©Monthly Notices of the Royal Astronomical Society
Earlier large-scale simulations largely omitted this dust physics. According to the Royal Astronomical Society, COLIBRE tracks three distinct grain chemistries, graphite, forsterite, and fayalite, in two size categories, and simulates how they grow, shatter, and are destroyed over time. The model required a clumping factor to compensate for the fact that the densest grain-growth environments, the cores of molecular clouds, remain too small to resolve even at COLIBRE’s scale.
The simulations run on the COSMA8 supercomputer at Durham University’s Institute for Computational Cosmology. The largest single run consumed 72 million CPU hours and used 136 billion particles, roughly 20 times more resolution elements than predecessor simulations like EAGLE, which the same team produced a decade earlier. “Much of the gas inside real galaxies is cold and dusty, but most previous large simulations had to ignore this,” said project leader Joop Schaye of Leiden University. “With COLIBRE, we finally bring these essential components into the picture.”
Validating the Standard Cosmological Model
The scientific significance of COLIBRE extends beyond technical novelty. When the James Webb Space Telescope began returning data on the early universe, some of its findings, particularly the apparent abundance of massive, bright galaxies at high redshift, were interpreted by some researchers as potential challenges to the Lambda Cold Dark Matter model, the standard framework of modern cosmology.
COLIBRE’s results provide a response. According to the published paper, the simulated galaxies reproduce observations from Webb and other instruments with strong numerical agreement across five orders of magnitude in stellar mass. The team found that once cold gas physics and dust are incorporated, the standard model remains consistent with what telescopes observe.
The L200m6 simulation zooms from the large-scale cosmic web down to individual galaxy structure, with gas temperature mapped across scales and stellar light rendered in Euclid colours accounting for dust. ©Monthly Notices of the Royal Astronomical Society
“Some early JWST results were thought to challenge the standard cosmological model,” said Evgenii Chaikin of Leiden University, lead author of several companion papers. “COLIBRE shows that, once key physical processes are represented more realistically, the model is consistent with what we see.”
The simulations are not without limitations. The so-called “Little Red Dots,” compact, red objects observed by Webb in large numbers around 600 million years after the Big Bang and absent shortly after, do not emerge from COLIBRE’s predictions. The team acknowledges that modeling these objects, which may represent early supermassive black hole seeds, will require higher resolution and new physics.
Core team member Carlos Frenk of Durham University described the broader result in terms of its perceptual impact on colleagues. He said that he routinely presents simulated galaxy images alongside real telescope data and asks observers to identify which is which, a test the synthetic images routinely pass.
Most COLIBRE simulations were completed in 2025. The team has also developed sonified video presentations of the data and interactive visualization tools, with the aim of making the results accessible beyond specialist audiences. Analysis of the data already produced is expected to continue for years.