Introducing the Lumina project: large-volume radiation-hydrodynamic simulations of the epochs of hydrogen and helium reionization

arXiv:2605.15310 · astro-ph.CO, astro-ph.GA · Submitted 2026-05-14 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Astrophysics Radio. Generated commentary on the latest astrophysics papers.

Vera: I'm Vera.

Jocelyn: And I'm Jocelyn, here with Subrahmanyan.

Vera: We're looking at the paper: "Introducing the Lumina project".

Jocelyn: Here's the gist of the paper. Lumina is a large-volume radiation-hydrodynamic simulation designed to self-consistently track the coupled evolution of galaxies, Active Galactic Nuclei (AGN), and the intergalactic medium (IGM) through hydrogen (H I), helium (He I), and helium II (He II) reionization down to redshift z=3.

Vera: The simulation spans a cosmological volume with a comoving side length of L box = 500 cMpc, utilizing a high resolution of 2 times 6000 cubed elements, corresponding to baryonic mass resolutions of 3.6 times 10 6 M and dark matter mass resolutions of 1.9 times 10 7 M.

Jocelyn: It employs the moving-mesh code *arepo* and integrates the IllustrisTNG galaxy-formation model with a GPU-accelerated M1 radiation-transport solver across six frequency bins.

Vera: The simulation incorporates sophisticated physical treatments, including replacing the equilibrium cooling treatment of TNG with a non-equilibrium primordial chemistry network for low-density gas (n H < 0.106 cm-3), and treating gas as optically thin above this threshold.

Title and authors: Vera: So, this paper, "Introducing the Lumina project: large-volume radiation-hydrodynamic simulations of the epochs of hydrogen and helium reionization", it sets up this simulation framework that covers a comoving side length of five hundred comoving megaparsecs.

Jocelyn: It’s not just tracking one thing; it’s trying to model the coupled evolution of galaxies, active galactic nuclei, and the intergalactic medium through hydrogen, helium I, and helium II reionization all at once.

Subrahmanyan: That coupling is what makes it interesting for cosmology because it lets them see how the sources driving these events—like black holes—actually influence the gas that's being ionized.

Vera: They use a high resolution of two by six thousand cubed elements, giving them baryonic mass resolutions of three point six times ten to the six solar masses and dark matter mass resolutions of one point nine times ten to the seven solar masses.

Jocelyn: That level of resolution is important for getting the right details on galaxy populations that feed those ionizing photons.

Subrahmanyan: And they’re using a moving mesh code called arepo, which lets them handle the complex dynamics of gas movement quite well across that large volume.

The paper's summary: Vera: The summary says they put sophisticated physical treatments in place, including replacing the equilibrium cooling treatment from the IllustrisTNG model with a non-equilibrium primordial chemistry network for low density gas.

Jocelyn: And treating the gas as optically thin above a certain density threshold, which is how they handle things at different scales within that simulation.

Subrahmanyan: They model radiation sources through three classes: stellar populations, high mass X-ray binaries and shocked hot interstellar gas, and AGN, which are what provide those hard photons driving helium II reionization.

Vera: They also incorporated a new initial conditions generator that self-consistently includes distinct baryon and dark matter transfer functions along with the relative baryon dark matter streaming velocity.

Jocelyn: That streaming velocity detail is important because it affects how fast star formation can start early on, which feeds into the whole reionization timeline.

Subrahmanyan: So, to put it simply, they’re using a combination of detailed chemistry networks and radiation transport to see how the different sources carve out those ionized bubbles over cosmic time.

The paper's improvements: Vera: The paper points out several improvements they made. First, they enhanced the modeling of reionization topology to self-consistently model the topology, timing, and thermal impact of each transition across hydrogen and helium reionization within that five hundred comoving megaparsec volume.

Jocelyn: That sounds like a big step because it means they aren't just looking at when things happen; they are looking at how the physical state of the gas changes during those transitions.

Subrahmanyan: They also improved how they link observables to these models by enabling the forward modeling of observables linking hydrogen and helium reionization topologies to evolving galaxy and AGN populations.

Vera: And they added a self-consistent initial conditions generator that includes coherent baryon dark matter streaming velocities in the initial conditions, which lets them assess their impact on early star formation and ionizing emissivity.

Jocelyn: Plus, they unified the treatment of radiation sources by allowing the AI to handle integrated flux from four distinct source classes: stellar populations, accreting black holes, high mass X-ray binaries, and shock heated interstellar medium.

Conclusion: Vera: So when we wrap up this discussion on "Introducing the Lumina project: large-volume radiation-hydrodynamic simulations of the epochs of hydrogen and helium reionization", it seems they’ve successfully connected the reionization topology, the IGM thermal history, and source demographics within a single framework.

Jocelyn: They’ve shown that hydrogen and helium reionization are jointly modeled, which is something that was previously harder to see as two separate events.

Subrahmanyan: This work demonstrates that they are jointly modeling these transitions, establishing it as a unique radiation hydrodynamics simulation combining large volume with coverage of both epochs.

Vera: They’ve made some specific predictions about the thermal history, noting a significant difference between runs: Lumina-fifty heats the neutral IGM to about two thousand seven hundred Kelvin by redshift five point seven, whereas Lumina-fifty NoX only reaches about eighty Kelvin.

Jocelyn: That difference in temperature is really crucial because it’s what we need to predict the two point one centimeter signal from the cosmic microwave background, which is tied to that early thermal state.

Subrahmanyan: I think the modeling of the source budget also helps tie everything together by showing that H I and He I reionization are driven by stellar emission globally, while AGN contribute only at a percent level around fourteen point at redshift five.

Vera: Overall, this simulation is powerful because it connects the topology of reionization to the actual physical state of the gas in a way that was hard to achieve before.

Jocelyn: It’s a lot of detail, but it gives us a much richer picture of how these two major epochs unfold.

Subrahmanyan: It sets up an important benchmark for future simulations looking at structure formation and how feedback from the smallest scales affects the largest cosmological volumes.

Oliver Zier, Aaron Smith, Xuejian Shen, Rongrong Liu, Rahul Kannan, Sonja M. Koehler, Volker Springel, Rüdiger Pakmor, Mark Vogelsberger, Teodora-Elena Bulichi and Lars Hernquist

Center for Astrophysics | Harvard & Smithsonian · Department of Physics, The University of Texas at Dallas Department of Physics, Massachusetts Institute of Technology Kavli Institute for Astrophysics and Space Research Max Planck Institute for Astrophysics

astro-ph.CO, astro-ph.GA

Submitted: 2026-05-14

Updated: 2026-08-03

Comments: 40 pages, 30 figures. Accepted by MNRAS, Please visit https://lumina-simulation.com for more details

Journal ref: Mon Not R Astron Soc (2026)

DOI: 10.1093/mnras/stag1469

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 99/100

The gist: Lumina is a large-volume radiation-hydrodynamic simulation designed to self-consistently track the coupled evolution of galaxies, Active Galactic Nuclei (AGN)…

Key concepts

Radiation-Hydrodynamic Simulation (RHD)
This is a complex computer model that tracks the movement of gas (hydrodynamics) while simultaneously calculating how radiation (photons) interacts with it. Lumina uses this to see how galaxies, gas, and radiation evolve together over cosmic time.
Hydrogen ($ ext{H I}$) Reionization
This is the process where neutral hydrogen gas in the early universe is ionized by ultraviolet light from first stars and galaxies. Lumina found that this happened late, driven mainly by stellar sources, with residual neutral patches persisting until $z ext{ approx } 4.75.
Helium II ($ ext{He II}$) Reionization
This is the ionization of helium from its two-electron state into a fully ionized state, which requires very hard photons. Lumina showed this process was driven by Active Galactic Nuclei (AGN) starting around $z ext{ approx } 6$ and was nearly complete by $z=3.

Terminology

Summary

Lumina is a large-volume radiation-hydrodynamic simulation designed to self-consistently track the coupled evolution of galaxies, Active Galactic Nuclei (AGN), and the intergalactic medium (IGM) through hydrogen (H I), helium (He I), and helium II (He II) reionization down to redshift z=3. p. 0 The simulation spans a cosmological volume with a comoving side length of L box = 500 cMpc, utilizing a high resolution of 2 times 6000 cubed elements, corresponding to baryonic mass resolutions of 3. p. 06 times 10 6 M and dark matter mass resolutions of 1. p. 09 times 10 7 M. p. 1 It employs the moving-mesh code arepo and integrates the IllustrisTNG galaxy-formation model with a GPU-accelerated M1 radiation-transport solver across six frequency bins.

The simulation incorporates sophisticated physical treatments, including replacing the equilibrium cooling treatment of TNG with a non-equilibrium primordial chemistry network for low-density gas (n H < 0. p. 6106 cm-3), and treating gas as optically thin above this threshold. p. 2 The radiation sources are modeled through three classes: stellar populations (using BPASS v2. p. 82.1 spectra with an effective escape fraction f esc = 0. p. 918), high-mass X-ray binaries and shocked, hot interstellar gas, and AGN, which provide the hard photons driving He II reionization. p. 8 The simulation also incorporates a new initial-conditions generator that self-consistently includes distinct baryon and dark matter transfer functions along with the relative baryon–dark matter streaming velocity. p. 0

Key findings regarding reionization history include:

  1. ** H I Reionization:** Lumina predicts a late, predominantly stellar-driven H I reionization. The median sub-volume is fully ionized by z about 5. p. 02, with residual neutral H I patches persisting until z about 4. p. 075. This late end to reionization is consistent with observational constraints derived from the high Ly alpha-forest opacity at z about 5. p. 05.

  2. ** He II Reionization:** He II reionization is driven self-consistently by AGN and begins around z 6. It is nearly complete by z=3, with a midpoint predicted near z 4 and a residual neutral He II volume filling fraction below one percent, aligning with current observational constraints. p. 0

  3. Thermal History: The IGM exhibits a multi-stage thermal history: early adiabatic cooling, X-ray pre-heating above the CMB temperature, rapid photoheating during H I and He I reionization (T (1–2) times 10 4 K), subsequent cooling, renewed heating during He II reionization, and late-time adiabatic cooling. Regions that reionize early in H I tend to reionize early in He II, leading to correlated temperature peaks. p. 1 A significant difference is noted between runs: Lumina-50 heats the neutral IGM to about 2700 K by z=5. p. 87, whereas Lumina-50 NoX only reaches about 80 K, a crucial distinction for predicting the 21 cm signal. p. 8

  4. Source Budget: Globally, H I and He I reionization are driven by stellar emission, while AGN contribute only at the percent level (around 14% at z=5). AGN dominate the hard photon budget for He II ionization below z 7. p. 1

  5. Galaxy Population: The galaxy population remains broadly consistent with IllustrisTNG expectations and observations, though Lumina shows slight deviations: it underpredicts the abundance of very massive galaxies (M* 10 11 M) at z about 5, and overpredicts stellar masses for haloes with M halo at least 10 13 M at z=3.

Lumina successfully connects reionization topology, the IGM thermal history, and source demographics within a single framework. p. 3 It demonstrates that H I and He II reionization are jointly modeled, establishing it as a unique RHD simulation combining large volume with coverage of both epochs. p. 0

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Enhanced modeling of reionization topology: The improved system can now self-consistently model the topology, timing, and thermal impact of each transition across hydrogen and helium reionization within a 500 cMpc volume.

  2. Precise source demographics linking to observables: The simulation enables the forward modelling of observables linking H i and He ii topologies to the evolving galaxy and AGN populations.

  3. Self-consistent initial conditions incorporating streaming velocities: The system now includes coherent baryon–CDM streaming velocities in the initial conditions, enabling a self-consistent assessment of their impact on early star formation and on the ionizing emissivity.

  4. Unified treatment of radiation sources: The improved AI can handle the integrated flux from four distinct source classes—stellar populations, accreting black holes, high-mass X-ray binaries, and shock-heated interstellar medium—with specific SEDs for each.

  5. Resolution across cosmic time: The system maintains a dynamic range required to model the galaxy population that dominates hydrogen reionization and the cosmological volume needed to sample the rare, bright quasars that drive He ii reionization.

Sources

Related papers