Lumina Rebuilds the Universe's First 2 Billion Years, Helium Reionization Included

For its first few hundred million years, the universe was filled with neutral gas that absorbed ultraviolet light. Then the first stars, galaxies and black holes switched on and broke those atoms apart, a long process astronomers call reionization. The James Webb Space Telescope now sees galaxies from that period directly. It is much harder to say how that light changed everything between the galaxies.

A new simulation called Lumina tries to answer that question at a scale no earlier project has reached. Researchers at the Center for Astrophysics | Harvard & Smithsonian built it, and the work is published in the Monthly Notices of the Royal Astronomical Society. Lumina recreates roughly the first 2 billion years of cosmic history inside a cube about 1.5 billion light-years on a side. The team calls it the largest cosmological radiation-hydrodynamics simulation ever produced. It is also the first such simulation to follow helium reionization as it happens while tracking how that process affects its surroundings.

Two reionizations, not one

Reionization is usually described as a single event, but there are really two. Hydrogen loses its single electron fairly easily, and ordinary starlight can do the job. Helium is harder. Removing its second electron, turning singly ionized helium (HeII) into fully ionized helium, takes much more energetic photons. According to Universe Today, helium makes up about 25 percent of the intergalactic medium. Fully ionizing it needed hard radiation from quasars, the active galactic nuclei (AGN) powered by supermassive black holes, whose activity peaked around the era known as Cosmic Noon.

Simulations have usually handled one of these processes or the other, often in small volumes. Small volumes are a real problem for helium reionization because quasars are rare. A simulated box has to be enormous before it contains enough of them to give a representative picture.

What's inside the box

The project paper on arXiv gives the details. Lumina models a region 500 comoving megaparsecs across. Comoving distance is measured on a grid that expands with the universe. The simulation uses 2×6000³ resolution elements, which is about 432 billion gas cells and dark matter particles combined. Each baryonic element carries about 3.6 million solar masses, and each dark matter particle about 19 million.

The physics comes in layers. The gas dynamics run on the AREPO moving-mesh code with the IllustrisTNG galaxy formation model, so stars, galaxies and AGN form and evolve on their own instead of being added by hand. On top of that sits a GPU-based radiation transport scheme using the M1 method. It tracks light in six frequency bins, which lets the simulation tell soft starlight from the hard photons that can ionize helium. The run continues down to redshift 3, when the universe was about 2 billion years old.

The Smithsonian release says Lumina covers a volume more than 100 times larger than earlier comparable projects. Universe Today reports that it ran on more than 15,000 GPUs on Oak Ridge National Laboratory's Frontier supercomputer.

What it predicts

For hydrogen, Lumina predicts that reionization was late and driven mainly by stars. In the paper's analysis, the median sub-volume of the simulation is fully ionized by about redshift 5.2. Some neutral patches survive until around redshift 4.75. In other words, the hydrogen transition still had not finished in every part of the universe more than a billion years after the Big Bang.

Helium follows its own timeline. As summarized by Universe Today, HeII reionization begins around redshift 6, reaches its midpoint near redshift 4, and is nearly complete by redshift 3. At that point the residual HeII filling fraction is below 1 percent, which the paper's authors, quoted by the outlet, say agrees with current observational constraints. The arXiv abstract says AGN drive this process and that the simulation shows a clear late-time thermal boost in the intergalactic gas associated with it. That heating is something observers can try to measure.

The simulation also passes a standard check. The paper reports that its Thomson optical depth is in excellent agreement with measurements from the Planck satellite. This quantity measures how much the free electrons released by reionization scatter the cosmic microwave background. A model that got the overall timing badly wrong would miss it.

Why It Matters

JWST has filled the early universe with observed galaxies, and the hard part now is interpretation. Which sources produced the ionizing photons? How quickly did the neutral gas disappear, and was the process even across the sky or patchy? A simulation is useful here only if it is large enough to be statistically representative and physically detailed enough to produce numbers telescopes can actually measure. Lumina aims to do both.

Lead author Oliver Zier, an ITC Fellow at the CfA, said in the Smithsonian release that JWST has made the early universe an exciting frontier. The simulation is meant to help interpret that data and observations from NASA's Nancy Grace Roman Space Telescope, which Universe Today reports launched about two months ago. The Smithsonian release says the model gives astronomers a tool for interpreting discoveries from both telescopes.

The predictions can be tested and could turn out wrong, which is what makes them useful. If observers find that hydrogen reionization finished much earlier than redshift 5, or that the intergalactic gas lacks the late-time thermal boost Lumina links to helium reionization, the model's physics will need revising. If the observations line up, Lumina gives a consistent account connecting stars, black holes and the gas between galaxies across the universe's first 2 billion years.

Either way, the field now has a model that covers both reionizations in a box large enough for the rare quasars that drove the second one. That makes it a strong baseline for the JWST and Roman data still to come.