"Lumina Simulation: The Largest-Ever Model of the Universe's First Two Billion Years"

The James Webb Space Telescope keeps showing us a young universe that looks older and more organized than our theories expected, and that is exactly the kind of embarrassment a good simulation is built to resolve. Researchers at the Center for Astrophysics | Harvard & Smithsonian have now published one: Lumina, described in Monthly Notices of the Royal Astronomical Society in 2026, is the largest cosmological radiation-hydrodynamics simulation ever created, recreating how ordinary matter, dark matter and radiation interacted during the universe's first two billion years.

The scale is what makes Lumina different from a prettier picture of the same physics. It covers more than 100 times the volume of previous comparable simulations, a cube roughly 1.5 billion light-years on each side, comparable to the scale today's most powerful telescopes can actually observe. That matters because the early universe was not a uniform fog: galaxies formed inside a cosmic web of filaments, and the light from the first stars transformed the gas between them in a process called reionization. Lumina is the first simulation of its kind to capture helium reionization as it happened, letting its effects immediately feed back into the evolving model, while still resolving the individual galaxies and luminous objects driving the change. For the first time, as co-author Aaron Smith of the University of Texas at Dallas put it, researchers can follow both hydrogen and helium changing into their modern form across an enormous volume without losing sight of the objects doing the transforming.

Building it took more than 15,000 graphics processing units running on Frontier, one of the world's fastest supercomputers, at Oak Ridge National Laboratory in Tennessee. Lead author Oliver Zier, an ITC fellow at the Center for Astrophysics, describes several years of work with plenty of ups and downs before the simulation's predictions came to life in video, and that timeline is worth keeping in mind. The most honest thing about a simulation this large is that it is a bet made visible: every assumption about dark matter, gas physics and radiation is translated into a detailed prediction that telescopes can then confirm or break. When observations from Webb and the upcoming Nancy Roman Space Telescope match Lumina, confidence in the standard picture grows. When they do not, the mismatch itself becomes the discovery, pointing at physics we have not yet written down.

The team designed Lumina as a platform rather than a single result, and that is where my own prediction comes in. Future studies are planned to use it to investigate how gases move between galaxies, how radiation from the big bang shaped cosmic evolution, and how regions of intense star formation developed. So here is a checkable claim: by 31 December 2027, at least one peer-reviewed study that uses Lumina simulation data will be published. I put that at about 60 percent. The case for it is straightforward, since the data already exist, the introducing paper is out, and the authors have named the follow-up questions. The case against it is the ordinary friction of science, since peer review is slow, the first follow-up analyses may take longer than a year to mature, and a team can always decide that its next paper needs more work before submission. Either outcome teaches something: a quick follow-up would show a large simulation paying off as shared infrastructure, while a delay would show how long even finished data takes to become new knowledge.

What I like most about Lumina is what it refuses to do. It does not claim to have seen the early universe, and it does not pretend that a model is a measurement. It gives astronomers a universe they can argue with, in the same size as the one their telescopes can see, and then waits for the sky to answer back. That is a good description of what science is supposed to feel like, and it is why a simulation built from equations can still be genuinely exciting to read about.