Webb and Roman Telescopes Power Massive Lumina Early Universe Simulation

At the Oak Ridge National Laboratory in Tennessee, more than 15,000 graphics processing units on the Frontier supercomputer have spent months calculating the dawn of time. The result is Lumina, a massive new cosmological simulation detailed in the Monthly Notices of the Royal Astronomical Society on October 26, 2026.[1][2][3]

The project models the first two billion years of cosmic history across a volume of space spanning 1.6 billion light-years, or 500 comoving megaparsecs. It tracks how dark matter, gas, and radiation interacted to form the earliest stars and galaxies with unprecedented resolution.[1][3]

Developed by researchers at the Massachusetts Institute of Technology, the Center for Astrophysics Harvard and Smithsonian, and partner institutions, Lumina provides a theoretical sandbox. Astronomers will use it to interpret the flood of high-resolution data arriving from the James Webb Space Telescope and the upcoming Nancy Grace Roman Space Telescope.[1][2]

"With JWST, the early universe has become an exciting new frontier in astrophysics," said Oliver Zier, an ITC Fellow at the Center for Astrophysics and lead author of the project's foundational paper. "To understand what we are seeing with this telescope, we need to translate our theories into detailed predictions."[2]

Modeling the cosmic dark ages

The simulation begins roughly 45 million years after the Big Bang. During this period, known as the cosmic dark ages, the universe was dominated by a dense, opaque fog of neutral hydrogen gas that blocked light from traveling freely across space.[1][5]

Over the simulated two billion years, Lumina tracks the emergence of the first luminous sources. As early stars and supermassive black holes ignited, their intense ultraviolet and X-ray radiation began to clear the hydrogen fog, driving a phase transition known as the Epoch of Reionization.[3][5]

"Lumina does everything from small-scale star and galaxy formation to large-scale re-ionization of hydrogen and helium," said Mark Vogelsberger, an MIT physics professor and study co-author. He noted that the simulation operates in a volume 100 times larger than previous best-in-class models.[1]

Previous models struggled to capture this transition accurately because of the conflicting scales involved. Modeling reionization requires resolving individual galaxies to predict their radiation output, while simultaneously simulating a volume large enough to capture the rare, ultra-bright quasars that drove the process.[3][5]

A dual reionization breakthrough

Lumina is the first cosmological radiation-hydrodynamics simulation to model both hydrogen and helium reionization simultaneously. The model demonstrates that hydrogen reionization was primarily driven by the collective ultraviolet light of early galaxies and stars.[1][5]

Helium reionization, however, required significantly more energy to strip the heavier atoms of their electrons. The simulation shows this secondary phase was fueled by early supermassive black holes in their active, accreting phases, known as active galactic nuclei.[1][3]

The team achieved this by combining the established IllustrisTNG galaxy-formation model with a new, GPU-accelerated radiation-transport solver. This allowed the simulation to track light in six different frequency bins as it propagated through the intergalactic medium down to a redshift of z=3.[3][5]

The simulation predicts a predominantly stellar-driven hydrogen reionization that concludes with the median sub-volume fully ionized by a redshift of 5.2. Residual neutral hydrogen patches persist slightly longer, while the AGN-driven helium reionization is nearly complete by a redshift of 3.[3]

The resulting data set contains nearly half a trillion resolution elements. Visualizations produced by the team show intricate, web-like filaments of dark matter intertwined with rapidly moving gas, punctuated by the expanding bubbles of ionized space where early galaxies formed.[4][5]

Measuring ionized bubbles

Researchers are already using Lumina to analyze the specific geometry of these ionized regions. In a September 2026 preprint, the team used a three-dimensional watershed method to measure the shapes and volumes of the expanding bubbles of ionized gas.[5]

They found that the characteristic radius of these bubbles grew from about one comoving megaparsec at a redshift greater than 10, to over 100 comoving megaparsecs by a redshift of 6. The bulk of the bubble population remained elongated, or prolate, through most of the reionization epoch.[5]

However, as the bubbles began to merge and percolate near the end of the transition, their shapes shifted. The simulation revealed a sharp peak of flattened, oblate structures associated with a small number of massive bubbles dominating the late-stage cosmic web.[5]

Understanding these three-dimensional shapes is crucial for interpreting real-world data. Observations often only sample two-dimensional slices or single lines of sight through the patchy intergalactic gas, meaning astronomers must extrapolate the true volume of the structures they detect.[5]

Preparing for Roman and Webb

The sheer scale of Lumina is designed specifically to match the observational capabilities of next-generation observatories. While JWST provides deep, narrow views of the early universe, the Roman Space Telescope will capture wide-field panoramas of the cosmos.[2][4]

"One of Lumina's major advantages is its sheer size," said Xuejian Shen, a researcher on the project. "It produces enormous populations of simulated galaxies, finally allowing us to make comparisons with the very large galaxy samples that upcoming surveys will observe."[1]

By comparing the simulated galaxy populations with actual telescope observations, astrophysicists can test the accuracy of their current cosmological models. If the real universe looks different from Lumina's predictions, scientists will know exactly where their theories need revision.[2]

The research team has made the simulation data available to the broader scientific community. Future applications will include mapping the large-scale structure of the early universe and predicting the gravitational-wave patterns produced by the first merging black holes.[1][5]