Powered by the Frontier supercomputer, the project models star and galaxy formation across a cube spanning roughly 1.63 billion light-years.
A massive computing project has recreated the early chapters of cosmic history, shedding light on the era when the first stars and galaxies ignited.
Frontier Supercomputer Powers 100,000 Cores for Cosmic Simulation
During peak moments, the simulation mobilized over 100,000 processor cores and roughly 15,000 graphics processors. The modeled sector represents a cube with sides measuring approximately 1.63 billion light-years, containing around 432 billion elements representing dark matter and gas.
According to the published details, the scope of this volume is roughly 100 times larger than the volume covered by previous simulations for comparable studies. The project aims to clarify how the universe emerged from the cosmic “dark ages,” which refers to the period before any luminous objects appeared. By tracing initial irregularities in matter distribution, the model demonstrates how those fluctuations eventually sparked star and galaxy formation, altering between-galaxy gas through radiation.
IllustrisTNG Software and Future Observations with the James Webb Space Telescope
Lumina runs on an advanced version of the IllustrisTNG software suite, augmented with radiation propagation modeling. This setup allows researchers to simultaneously track gravity, matter movement, star creation, and the ionization of intergalactic gas. The model specifically highlights the reionization of hydrogen driven by light from young stars and galaxies, alongside helium ionization influenced by active supermassive black holes.
A key advantage of the project is its ability to connect individual galaxy development with shifts in the large-scale structure of the cosmos. Researchers plan to compare these simulation outputs against observations from the James Webb Space Telescope and the future Roman telescope to test theories of early star formation. While the simulation relies on physical models and initial conditions rather than serving as a direct recording of the past, the complete dataset is scheduled to become fully available to the scientific community in 2028.
