Weak Evolution of Cosmic Atomic Hydrogen over the Past 4.5 Billion Years

arXiv:2607.05326 · astro-ph.GA, astro-ph.CO · Submitted 2026-07-06 · 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: Next we'll be talking about the paper "Weak Evolution of Cosmic Atomic Hydrogen over the Past 4.5 Billion Years".

Jocelyn: The paper was written by the authors from National Natural Science Foundation of China and National Key Research and Development Program of China and National SKA Program of China and China Manned Space Project and Guizhou Provincial Science and Technology Projects and CAS Project for Young Scientists in Basic Research and Office of Science and Technology, Shanghai Municipal Government and National Astronomical Observatories of Chinese Academy of Sciences (NAOC) and Lawrence Berkeley National Laboratory and U.S. Department of Energy, Office of Science, Office of High-Energy Physics and DOE Office of Science User Facility and U.S. National Science Foundation (NSF) and National Optical-Infrared Astronomy Research Laboratory and Science and Technology Facilities Council of the United Kingdom and Gordon and Betty Moore Foundation and Heising-Simons Foundation and French Alternative Energies and Atomic Energy Commission (CEA) and National Council of Humanities, Science and Technology of Mexico (CONAHCYT) and Ministry of Science and Innovation of Spain (MICINN).

Vera: Stay tuned as we take you through the paper and discuss its implications.

Jocelyn: We also have Subrahmanyan with us today — guest researcher.

Vera: Alright, let's get started.

Title and Authors: Vera: We're starting today with a massive new paper titled "Weak Evolution of Cosmic Atomic Hydrogen over the Past four point five Billion Years."

Jocelyn: The title alone makes me think we're looking at a very steady, almost quiet period of cosmic history, Vera.

Vera: It really does, especially considering this is a huge effort by the FASHI collaboration, with lead authors like Chuan-Peng Zhang and Hong Guo.

Jocelyn: That's a lot of names, so I assume the data set they're using is just as enormous?

Subrahmanyan: It has to be, Jocelyn, because to claim something is "weakly evolving" across billions of years, you need a statistical sample that covers more than just a few local galaxies.

Vera: You're spot on, Subrahmanyan, because they've managed to pull together data for about two and a half million galaxies.

Jocelyn: Two and a half million? That sounds like an observationalist's dream, but what are they actually looking for in all that noise?

Subrahmanyan: They are essentially trying to track the baryon cycle, which is the process of gas falling into galaxies, turning into stars, and then being blown back out.

Vera: And specifically, they want to see if the reservoir of neutral atomic hydrogen, the H I, is drying up as the universe ages.

Jocelyn: So if the star formation rate is dropping, they're checking if it's because the galaxies are actually running out of H I gas.

Subrahmanyan: That's the fundamental question, and if the H I density stays relatively constant while stars stop forming, it tells us the "fuel" is there, but something is preventing it from being used.

Vera: It's a clever way to frame the whole problem, looking at the supply versus the consumption.

Jocelyn: I'm curious to see if the data actually shows that decoupling between the gas and the stars.

Vera: We'll get into those specific numbers in just a moment, but first, let's look at what their summary actually reveals.

Summary and Results: Vera: Moving into the results of "Weak Evolution of Cosmic Atomic Hydrogen over the Past four point five Billion Years," the numbers are actually quite surprising.

Jocelyn: Surprising in what way, Vera? Did they find that the gas is much more abundant than we thought?

Vera: Not necessarily more abundant, but it's much more stable; they saw the cosmic H I density, or HI, decrease by a factor of only about one point three five over the last four point five billion years.

Jocelyn: That seems like a very small change compared to the massive drop we see in star formation.

Subrahmanyan: That's exactly the point, Jocelyn, because the cosmic star formation rate density has actually declined by a factor of about two point four six in that same period.

Vera: It's a huge discrepancy, isn't it? The stars are slowing down significantly, but the atomic hydrogen isn't disappearing at the same rate.

Jocelyn: So the H I isn't the limiting factor for star formation at late times?

Subrahmanyan: Precisely, and this suggests that the bottleneck is likely the conversion of that atomic hydrogen into molecular hydrogen, which is what stars actually form from.

Vera: They even looked at the gas fraction within individual galaxies, which they call f HI, and found it only changes by less than zero point two dex across different stellar masses.

Jocelyn: That means even the massive galaxies are holding onto their atomic gas in a very consistent way?

Vera: Yes, the relation between mass and gas fraction looks remarkably similar regardless of the redshift they're looking at.

Subrahmanyan: This is a massive constraint for our theoretical models, because it means the baryon cycle is incredibly well-regulated.

Jocelyn: It sounds like the universe has a built-in thermostat for its gas supply.

Vera: It really does, and it forces us to rethink how gas accretion and feedback processes work together to keep things steady.

Jocelyn: I want to know how they pulled off such a precise measurement without getting lost in the radio interference.

Methodology and Improvements: Vera: To understand how they got those results in "Weak Evolution of Cosmic Atomic Hydrogen over the Past four point five Billion Years," we have to talk about the synergy between FAST and DESI.

Jocelyn: I've heard about the FAST telescope, but how does it work with an optical survey like DESI?

Vera: It's a brilliant combination; they use the incredible sensitivity of the Five-hundred-meter Aperture Spherical Telescope for the radio signals and then use DESI to get incredibly precise redshifts for millions of galaxies.

Jocelyn: But the H I signal from a single galaxy at that distance must be incredibly faint, right?

Subrahmanyan: It's far too faint to detect individually with this setup, which is why they use a technique called spectral stacking.

Vera: Exactly, they take the spectra from thousands of similar galaxies and co-add them to boost the signal-to-noise ratio until the H I line finally pops out.

Jocelyn: That sounds like it would be prone to all sorts of errors, especially with radio frequency interference.

Subrahmanyan: They had to be extremely careful with that, and they even used a "confusion correction" to make sure they weren't accidentally picking up signal from a nearby massive galaxy.

Vera: They also had to deal with luminosity bias, where the survey naturally picks up the brightest, gas-rich galaxies more easily than the faint ones.

Jocelyn: So they didn't just take the data at face value; they had to build a mathematical model to correct for those selection effects?

Vera: They did, using a very sophisticated approach to ensure their measurements of the gas fraction were truly representative of the whole population.

Subrahmanyan: This level of rigor is what allows them to turn a potentially messy dataset into a high-precision benchmark for galaxy evolution.

Jocelyn: It's impressive how they've turned the limitations of the telescopes into a controlled statistical experiment.

Vera: It really is, and it sets a very high bar for future radio surveys.

Conclusion: Vera: As we wrap up our look at "Weak Evolution of Cosmic Atomic Hydrogen over the Past four point five Billion Years," it's clear this paper changes the conversation.

Jocelyn: It really does, because it shifts the focus from "where did the gas go?" to "why isn't the gas turning into stars?"

Subrahmanyan: It's a subtle but profound distinction that moves us toward a much more nuanced understanding of the baryon cycle.

Vera: The fact that HI remains so stable while star formation plummets is just such a clean, powerful result.

Jocelyn: I'm walking away thinking that the atomic hydrogen reservoir is much more resilient than we gave it credit for.

Subrahmanyan: And for theorists, it means our simulations have to be much better at modeling the transition from atomic to molecular phases to match this stability.

Vera: It's a massive win for the FASHI collaboration and a great example of what happens when you combine radio and optical data.

Jocelyn: I can't wait to see if the next generation of surveys, like the ones coming from SKA, confirm this level of precision.

Subrahmanyan: We'll be watching the data very closely, because that's where the next big piece of the puzzle is hiding.

Vera: Well, it's been a pleasure digging into these details with you both today.

Jocelyn: I'm already looking forward to the next paper.

Subrahmanyan: Just keep questioning the noise, and the signal will follow.

Vera: We'll be back after the break to look at a paper concerning galactic magnetic fields.

National Natural Science Foundation of China · National Key Research and Development Program of China · National SKA Program of China · China Manned Space Project · Guizhou Provincial Science and Technology Projects · CAS Project for Young Scientists in Basic Research · Office of Science and Technology, Shanghai Municipal Government · National Astronomical Observatories of Chinese Academy of Sciences (NAOC) · Lawrence Berkeley National Laboratory · U.S. Department of Energy, Office of Science, Office of High-Energy Physics · DOE Office of Science User Facility · U.S. National Science Foundation (NSF) · National Optical-Infrared Astronomy Research Laboratory · Science and Technology Facilities Council of the United Kingdom · Gordon and Betty Moore Foundation · Heising-Simons Foundation · French Alternative Energies and Atomic Energy Commission (CEA) · National Council of Humanities, Science and Technology of Mexico (CONAHCYT) · Ministry of Science and Innovation of Spain (MICINN)

astro-ph.GA, astro-ph.CO

Submitted: 2026-07-06

Updated: 2026-07-06

Comments: Accepted in principle by Nature Astronomy. This is the authors' manuscript version

Code: https://github.com/desihub/fiberassign

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 86/100

The gist: The provided text excerpts contain sections detailing references, acknowledgments, author contributions, and data availability.

Key concepts

Cosmic Atomic Hydrogen Density (Ω_HI)
This measures the overall density of neutral atomic hydrogen gas across the universe over time. The study found that this density decreased only slightly over 4.5 billion years, suggesting that the cosmic supply of H I gas is highly stable.
Baryon Cycle
This describes the continuous process where gas falls into galaxies, forms stars, and then gets blown back out into space. Researchers use this cycle to determine if galaxies are running out of available fuel (H I) as the universe ages.
Spectral Stacking
Since the H I signal from a single distant galaxy is too faint to detect, researchers employ spectral stacking. This technique involves collecting and mathematically adding together spectra from thousands of similar galaxies to boost the overall detectable signal.

Terminology

Summary

The provided text excerpts contain sections detailing references, acknowledgments, author contributions, and data availability. They do not include the abstract or scientific summary of the paper Weak Evolution of Cosmic Atomic Hydrogen over the Past 4.5 Billion Years. Therefore, I cannot extract a summary of the scientific findings from these pages.

Improvements for AI systems

(Self-Correction Note: Given the source material is highly specialized astrophysics literature concerning large surveys, cosmological simulations, and multi-wavelength gas physics, the AI improvements must move beyond general ML tasks and focus on domain-specific scientific inference and data handling.)

The core challenge revealed by this bibliography is the synthesis of massive, multi-physics datasets (simulations) with sparse, noisy observational data (DESI/FAST spectra). Current AI systems often fail when faced with highly non-stationary noise profiles or when physical causality must be inferred from correlated but non-causal observational metrics.

Here are four specific, high-impact improvements for AI systems derived from the scientific challenges presented:

Target Scientific Problem: Analyzing the vast parameter space of cosmological simulations (e.g., TNG50, IllustrisTNG) and inferring physical processes (like the atomic-to-molecular gas transition [92]) that are computationally expensive to resolve fully.

The Improvement: Develop a Conditional Variational Autoencoder (CVAE) architecture trained not just on the input state vector (X, e.g., density, temperature, metallicity) but conditioned explicitly on known physical laws and equations of state (L physics). The latent space must be regularized using a physics loss term L physics = E[grad times (rho v) +].

What the Improved AI System Can Do:

  • Accelerated Parameter Mapping: It can generate high-fidelity, physically consistent alternate realizations of gas structures and galaxy evolution tracks (e.g., simulating the impact of varying feedback coupling efficiencies) orders of magnitude faster than running full hydrodynamic simulations.

  • Feature Extraction in Latent Space: Instead of simply reproducing a state, the system can navigate the latent space to identify the minimum necessary changes in initial conditions or physical parameters required to explain a specific observed discrepancy (e.g., determining if stellar mass quenching is better modeled by environmental stripping or internal feedback).

  • Quantifying Uncertainty: It provides not just a prediction, but a quantifiable measure of how far that prediction lies from the established boundaries of known physics, which is crucial for high-stakes scientific inference.

Target Scientific Problem: Analyzing extremely faint spectral features (like HI gas profiles) in massive surveys (DESI/FAST) where the signal-to-noise ratio (SNR) drops drastically, and the noise profile is non-Gaussian and time/space-dependent.

Target Scientific Problem: Disentangling the dominant physical mechanisms responsible for galaxy evolution—specifically, whether stellar mass quenching is primarily driven by internal processes (e.g., AGN feedback) or external environmental effects (e.g., ram pressure stripping/galaxy mergers [85], [93]).

Target Scientific Problem: Synthesizing knowledge across disparate, highly specialized sources (

Abstract

The cosmic star formation rate density (CSFRD) has declined sharply toward the present day, but the roles of the atomic and molecular gas reservoirs remain uncertain. We measure the cosmic HI density, HI, over 0<z<0.41 by combining HI spectra from the Five-hundred-meter Aperture Spherical Telescope with optical spectroscopy from the Dark Energy Spectroscopic Instrument for about2.5 million galaxies across about12,000, deg squared. We measure a raw decrease in HI by a factor of 1.35 plus or minus0.10 over the past 4.5 Gyr. Even after applying the conservative systematic corrections from our forward model, the inferred decline is only 1.12 plus or minus0.10 -- still far weaker than the CSFRD decline (a factor of 2.46). The molecular gas density, in contrast, is known to evolve more closely with star formation. At fixed stellar mass, the average HI gas fraction evolves by less than 0.2 dex, showing that the weak evolution is present across the galaxy population. These quantitative differences rule out rapid depletion of galaxy HI as the primary driver of the late-time CSFRD decline, and provide a stringent benchmark for models of gas accretion, phase conversion and star-formation regulation.

Sources

Related papers