🔭 astrophysics

The impact of source and survey modelling on the connection between [O III] emitters and Ly $\alpha$ forest transmission at z ~ 6

This paper presents an empirical model incorporating JWST survey geometry to show that while large scatter in mock galaxy-Ly $\alpha$ transmission cross-correlations allows current observations to be consistent with various ionising source models, the significant discrepancy in correlation peak scales and current data limitations highlight the need for larger sample sizes and larger simulation volumes to further constrain the connection between [O III] emitters and the intergalactic medium at z ~ 6.

Original authors: Luke Conaboy, James S. Bolton, Laura C. Keating, Martin G. Haehnelt, Girish Kulkarni, Ewald Puchwein

Published 2026-06-03

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Luke Conaboy, James S. Bolton, Laura C. Keating, Martin G. Haehnelt, Girish Kulkarni, Ewald Puchwein

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: A Cosmic Game of "Where's Waldo?"

Imagine the early universe (about 13 billion years ago) as a giant, foggy room. This "fog" was made of neutral hydrogen gas that blocked light. Over time, the first stars and galaxies turned on like lightbulbs, burning away the fog and making the room transparent. This process is called reionization.

Scientists want to know: Who turned on the lights? Was it just the biggest, brightest galaxies, or did thousands of tiny, dim ones do the work too?

To find out, astronomers are using the James Webb Space Telescope (JWST) to look for specific glowing galaxies (called [O III] emitters) near high-redshift quasars (super-bright beacons). They are checking if the "fog" (hydrogen gas) is thinner or thicker near these galaxies.

The Problem: The Map Doesn't Match the Territory

Previous computer models tried to predict what we should see. They said, "If galaxies are the lightbulbs, the fog should clear up in a specific pattern around them."

However, recent real-world observations showed a pattern that didn't quite match the old models. The "clearing" of the fog seemed to happen at a different distance from the galaxies than the models predicted. It was like looking at a map of a city and finding that the traffic jams were happening on the wrong streets.

What This Paper Did: Building a Better Simulator

The authors of this paper decided to build a much more realistic "video game" simulation to see if the mismatch was real or just a flaw in how they were playing the game.

1. The "Cast of Characters" (Source Modelling)
In the old models, they picked galaxies somewhat randomly, like grabbing a handful of marbles from a jar. In this new study, they used a technique called Abundance Matching.

The Analogy: Imagine you have a jar of marbles of different sizes (representing dark matter halos). Instead of just picking the big ones, they carefully matched the number of marbles to the number of actual glowing galaxies we see in the sky. They made sure their simulation had the right mix of bright and faint galaxies, just like a real census.

2. The "Camera Angle" (Survey Modelling)
This is the most important part. The old models looked at the entire universe in the simulation. But JWST doesn't see the whole universe; it sees a tiny slice, like looking through a straw.

The Analogy: Imagine trying to guess the weather in a whole country by only looking out your kitchen window. If you only look at your window, you might see rain, but the rest of the country could be sunny.
The authors built "mock surveys" that mimicked exactly how JWST looks at the sky: the size of the window, the depth of the view, and the specific galaxies it can actually detect. They ran this simulation 1,024 times to see how much the results change just by chance.

The Results: It's All About the Scatter

When they compared their new, fancy simulation to the real JWST data, they found two main things:

1. The "Peak" is Still Off
The simulation still showed the "fog clearing" happening closer to the galaxies (about 20 million light-years away) than the real data suggests (about 30 million light-years away).

The Takeaway: Even with a better model of which galaxies are present, the simulation still can't perfectly explain why the clearing happens so far away. The physics might still be missing something, or the galaxies might be living in slightly heavier "homes" (dark matter halos) than we thought.

2. The "Noise" is Huge (The Scatter)
This is the biggest discovery. Because the universe is so vast and the JWST view is so small, the results vary wildly from one "mock survey" to the next.

The Analogy: Imagine trying to guess the average height of people in a city by measuring just five people standing on a street corner. Sometimes you pick five basketball players (tall average); sometimes you pick five jockeys (short average).
The authors found that the "noise" or scatter in their data is so large that almost any model can fit the real data. If you look at one random slice of the universe, the data might look like Model A. If you look at another slice, it might look like Model B.

The Conclusion: We Need More Data

The paper concludes that right now, the "fog" is too thick (metaphorically speaking) to tell the difference between different theories. The large scatter means that current observations can't rule out many different ideas about what powered the reionization of the universe.

The Verdict: The mismatch between the model and the data might just be a fluke of looking at a small sample size.
The Fix: To solve this, we need more data (bigger sample sizes from JWST) and bigger simulations (larger virtual universes) to smooth out the noise and see the true pattern.

In short: The authors built a better camera and a better map, but they realized that because the universe is so big and we are only looking at a tiny corner of it, the picture is still very fuzzy. We need to take more photos to get a clear focus.

Technical Summary: The Impact of Source and Survey Modelling on the Connection Between [O iii] Emitters and Ly $\alpha$ Forest Transmission at $z \sim 6$

Problem Statement
Recent observations from the James Webb Space Telescope (JWST) of [O iii]-emitting galaxies have revealed a cross-correlation with Ly $\alpha$ forest transmission that presents a challenge to existing numerical models. Specifically, measurements indicate an excess in intergalactic Ly $\alpha$ transmission at scales of $\sim 10\text{--}40$ cMpc from galaxies. While theoretical models predict such an excess driven by enhanced ionization around galaxies, reproducing the observed scale and amplitude has proven difficult. Previous theoretical work often lacked detailed modeling of the [O iii] emitter population or the specific geometric constraints and depth of JWST surveys (such as ASPIRE and EIGER), potentially leading to discrepancies between simulation and observation.

Methodology
The authors utilize the Sherwood-Relics suite of high-resolution cosmological hydrodynamical simulations, which employ a hybrid radiative transfer scheme to model patchy reionization and complete reionization at $z=5.3$ . To bridge the gap between simulation haloes and observed galaxies, the authors construct a new empirical model incorporating the following steps:

Abundance Matching: They map the simulation's dark matter halo mass function to the observed UV luminosity function of galaxies at $z \sim 6$ (based on Bouwens et al. 2021) using the abundance matching technique.
Luminosity Conversion: UV luminosities are converted to [O iii] luminosities using a scaling relation fitted to JWST data (Matthee et al. 2023). This relation accounts for the scatter in the $L_{[\text{O III}]}/L_{\text{UV}}$ ratio and applies a limiting flux sensitivity ( $1.0 \times 10^{-18}$ erg s $^{-1}$ cm $^{-2}$ ) to mimic survey detection limits.
Survey Forward Modeling: The authors generate mock survey catalogues that explicitly incorporate the geometry and depth of the ASPIRE and EIGER surveys. This includes simulating specific quasar sightlines, field-of-view constraints, and the number of detected emitters per sightline.
Statistical Analysis: They generate an ensemble of 1,024 realizations for ASPIRE-like surveys and 2,500 realizations for Zhu et al. (2026)-like surveys to quantify the variance in cross-correlation measurements. They compare these mocks against observational data for:
- Galaxy clustering (volume-averaged projected autocorrelation).
- One-dimensional galaxy–Ly $\alpha$ transmission cross-correlation.
- Two-dimensional galaxy surface density vs. Ly $\alpha$ effective optical depth ( $\tau_{\text{eff}}^{50}$ ).

Key Results

Galaxy Clustering: The model reproduces the observed clustering of [O iii] emitters (Huang et al. 2026; Eilers et al. 2024) with reasonable agreement ( $< 1\text{--}2\sigma$ ) when applying a luminosity cut of $\log_{10}(L_{[\text{O III}]}/\text{erg s}^{-1}) \ge 42.4$ . Discrepancies at small scales ( $r_\perp \lesssim 0.4$ cMpc) are attributed to the "one-halo term" missing in the simulation's Friends-of-Friends halo finder, which cannot host multiple galaxies within a single halo.
One-Dimensional Cross-Correlation: The empirical source and survey modeling leads to a slight increase in the mean host halo mass ( $\langle \log_{10}(M_h/M_\odot) \rangle \approx 10.7$ ) compared to previous simpler mass-cut approaches. However, this does not resolve the scale discrepancy: the peak of the cross-correlation in the mocks occurs at $\sim 20$ cMpc, whereas observations show a peak at $\sim 30$ cMpc.
Large Scatter: A primary finding is the very large scatter in the mock survey measurements of the cross-correlation, particularly at higher redshifts ( $z=6.4$ ). This scatter arises from the rarity of Ly $\alpha$ transmission spikes and the limited number of sightlines in actual surveys. Consequently, the large variance allows the model to achieve a statistically good match to the observational data (including the high amplitude seen in Kakiichi et al. 2025) despite the mismatch in the peak location.
Two-Dimensional Cross-Correlation: The model shows good statistical agreement with the relationship between galaxy surface density and Ly $\alpha$ $α$ effective optical depth (Zhu et al. 2026). The results indicate that both the most transparent ( $\tau_{\text{eff}}^{50} \lesssim 3$ $τ_{eff}^{50} ≲ 3$ ) and most opaque ( $\tau_{\text{eff}}^{50} \gtrsim 5$ $τ_{eff}^{50} ≳ 5$ ) sightlines can coincide with galaxy underdensities.
- Transparent sightlines correspond to underdense, hot regions that have recently been reionized.
- Opaque sightlines correspond to colder regions with a higher likelihood of containing neutral islands ( $x_{\text{HI}} > 0.5$ over $\ge 2$ cMpc).

Significance and Conclusions
The paper argues that the large scatter inherent in current galaxy-IGM observations, driven by limited sample sizes and survey geometry, implies that present data may struggle to rule out a broad range of ionizing source models. While the detailed modeling of source selection and survey geometry is critical for accurately assessing measurement variance, it does not fundamentally alter the shape of the cross-correlation or resolve the scale discrepancy between models and observations.

The authors conclude that further progress requires:

Increased observational sample sizes from ongoing programs like EIGER and ASPIRE.
Simulations in larger volumes ( $> 250$ cMpc) to better capture the high-mass end of the halo mass function and reduce cosmic variance.
Diverse source models, including faint active galactic nuclei, to test the robustness of the reionization history constraints.

The study reinforces that the relationship between galaxies and the intergalactic medium at the tail-end of reionization is complex, where galaxy density does not simply scale with the ionization state of the gas.

The Big Picture: A Cosmic Game of "Where's Waldo?"

The Problem: The Map Doesn't Match the Territory

What This Paper Did: Building a Better Simulator

The Results: It's All About the Scatter

The Conclusion: We Need More Data

More like this