🔭 astrophysics

A Measurement of the Thermal and Ionization State of the IGM at $z < 0.5$

Using machine learning on Hubble Space Telescope quasar spectra, this study reveals that the low-redshift intergalactic medium is unexpectedly hot and nearly isothermal with a suppressed photoionization rate, suggesting that either new heating mechanisms or unresolved turbulence are required to explain the observed broad Lyman-alpha lines.

Original authors: Teng Hu, Vikram Khaire, Joseph F. Hennawi, Todd M. Tripp, Jose Onorbe, Michael Walther, Zarija Lukic

Published 2026-06-04

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Teng Hu, Vikram Khaire, Joseph F. Hennawi, Todd M. Tripp, Jose Onorbe, Michael Walther, Zarija Lukic

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: The Universe's "Empty" Space is Actually Hot

Imagine the universe as a giant, mostly empty room. Inside this room, there are a few pieces of furniture (stars and galaxies), but the vast majority of the space is filled with a very thin, invisible gas called the Intergalactic Medium (IGM). This gas is the largest reservoir of "stuff" (baryons) in the universe.

For a long time, scientists believed that after the universe went through a few major "heating events" (like turning on a cosmic oven to ionize helium), this gas would slowly cool down, much like a cup of coffee left on a table. By the time the universe reached its current age (which corresponds to a redshift of less than 0.5, or the last 5 billion years), they expected this gas to be quite chilly—around 4,000 Kelvin (about 3,700°C).

The Surprise:
This paper says: "Nope, it's not cool. It's scorching."

Using data from the Hubble Space Telescope, the authors found that this cosmic gas is actually extremely hot—about 28,000 Kelvin (roughly 25,000°C). That is seven times hotter than the standard theory predicted. It's as if you expected a cup of coffee to be lukewarm, but when you took a sip, it was boiling hot.

How They Measured It: The "Fingerprint" of Gas

The scientists couldn't just stick a thermometer in space. Instead, they looked at the "fingerprints" left by the gas.

The Light Source: They used light from distant, bright beacons called Quasars.
The Absorption: As this light travels through the universe to reach us, the thin gas in the IGM absorbs some of it, creating dark lines in the spectrum (like a barcode).
The Width of the Lines: The key clue is the width of these dark lines.
- If the gas is cold and calm, the lines are narrow.
- If the gas is hot and moving fast, the lines get wider (this is called "thermal broadening").
- Analogy: Think of a crowd of people. If they are standing still (cold), they look like a tight group. If they are running around excitedly (hot), they spread out and look blurry.

The authors analyzed 82 high-quality spectra from the Hubble Space Telescope. They used a sophisticated machine-learning system to measure the width of these lines and the amount of gas present.

The Results: A Mystery Heat Source

The team measured the temperature and the "ionization state" (how stripped of electrons the gas is) in four different time periods (redshifts) between now and 5 billion years ago.

Temperature: They found the gas is nearly isothermal, meaning it has a uniform temperature everywhere, rather than having hot and cold spots. The temperature is roughly 28,000 K.
The Ionization Rate: They also measured how much ultraviolet light from the rest of the universe is hitting this gas. They found this rate is lower than what current computer models predict.

The Conflict:
Standard computer models of the universe say the gas should have cooled down to ~4,000 K. The data says it's ~28,000 K. This is a 7-sigma discrepancy, which in science is a massive, undeniable gap. It's like a weather forecast predicting a blizzard, but you step outside and it's a heatwave.

Why Is It So Hot? (The Suspects)

Since the gas is hotter than expected, the authors investigated two main possibilities:

1. A Hidden Heating Mechanism (The "New Heater")
Maybe there is a heat source we don't know about that has been warming the gas for billions of years. The paper suggests several suspects:

Dark Matter: Maybe dark matter particles are colliding or decaying, releasing heat.
Gamma Rays: High-energy radiation from black holes (blazars) might be heating the gas.
Dust: Tiny dust grains floating in the void might be absorbing starlight and re-radiating heat.
Galaxy Feedback: Explosions from forming galaxies might be pumping energy into the gas, but current simulations aren't capturing this well.

2. Invisible Turbulence (The "Shaking Gas")
Maybe the gas isn't actually hotter; maybe it's just shaking.

Analogy: Imagine a calm pond. If you blow on it, the water ripples. If you shake the pond violently, the water looks "wide" and blurry, just like hot gas.
The authors tested if small-scale turbulence (gas moving chaotically at speeds of about 15 km/s) could explain the wide lines without needing extra heat. They found that if you add this specific amount of "shaking" to the standard cold model, it matches the observations. However, this turbulence would need to be getting stronger as the universe gets older, which is also a mystery.

Ruling Out the "Broken Camera"

Before declaring the universe is broken, you have to check if your camera is broken. The authors worried that maybe the Hubble telescope's resolution was overestimated, making the lines look wider than they really are.

The Test: They compared Hubble's data with data from the STIS instrument, which has much higher resolution (like comparing a standard definition TV to a 4K TV).
The Verdict: The high-resolution data matched the Hubble data perfectly when Hubble's known blur was applied. This proves the Hubble instrument is working correctly. The "blur" is real, not an error.

Conclusion

The universe's "empty" space is not cooling down as we thought. It is unexpectedly hot and uniform.

If it's heat: We need to find a new cosmic heater (like exotic dark matter or dust).
If it's motion: We need to find a source of invisible turbulence that is getting stronger over time.

Either way, our current understanding of how the universe cools down after its early "hot phase" is incomplete, and this paper highlights a major gap in our knowledge of the last 5 billion years of cosmic history.

Technical Summary: A Measurement of the Thermal and Ionization State of the IGM at $z < 0.5$

Problem Statement
The thermal and ionization state of the Intergalactic Medium (IGM) at low redshift ( $z < 0.5$ ) presents a significant discrepancy between observational data and theoretical predictions. While high-redshift IGM models align well with Ly $\alpha$ forest observations, low-redshift data from the Hubble Space Telescope (HST) Cosmic Origins Spectrograph (COS) reveal Ly $\alpha$ absorption lines that are significantly broader than predicted. Standard cosmological models, which account for adiabatic cooling following He II reionization ( $z \approx 3$ ), predict a mean IGM temperature ( $T_0$ ) of $\sim 4000$ K and a temperature-density slope ( $\gamma$ ) of $\sim 1.6$ at $z=0$ . However, previous studies have suggested the low- $z$ IGM may be hotter or more turbulent than these canonical values, yet precise measurements have been hindered by limited data sizes and the degeneracy between thermal state and the hydrogen photoionization rate ( $\Gamma_{\text{HI}}$ ).

Methodology
This study employs a machine-learning-based inference framework to measure the thermal state ( $T_0, \gamma$ ) and ionization state ( $\Gamma_{\text{HI}}$ ) of the IGM across four redshift bins: $z=0.1, 0.2, 0.3,$ and $0.4$.

Data Processing: The analysis utilizes 82 archival HST/COS quasar spectra (Danforth et al. 2016) with high signal-to-noise ratios. The authors perform automated Voigt profile fitting using a custom wrapper for the VPFIT code to extract the joint Doppler parameter ( $b$ ) and neutral hydrogen column density ( $N_{\text{HI}}$ ) distribution. Metal absorbers and proximity zones are masked based on identifications from Danforth et al. (2016), resulting in a final sample of 657 Ly $\alpha$ absorption lines.
Simulations and Forward Modeling: The authors use a suite of 51 Nyx cosmological hydrodynamic simulations with varying thermal histories. To address the difficulty of generating high-temperature models at low redshift via photoheating alone, they apply a post-processing temperature rescaling technique to generate models with $T_0$ up to $\sim 30,000$ K. They also vary $\Gamma_{\text{HI}}$ across 13 values. Crucially, the forward modeling includes the specific, non-Gaussian Line Spread Function (LSF) of the COS instrument and adds a noise floor to match observational artifacts.
Inference Engine: The study utilizes Density Estimation Likelihood-Free Inference (DELFI) with neural density estimators. A neural network is trained to learn the conditional probability distribution $P(b, N_{\text{HI}} | T_0, \gamma, \Gamma_{\text{HI}})$ from the simulated mock spectra. This emulator, combined with a Gaussian process emulator for the absorber line density ($dN/dz$), allows for Bayesian inference of the IGM parameters directly from the observed $b-N_{\text{HI}}$ distribution and line counts, effectively breaking the degeneracy between thermal state and ionization rate.

Key Results
The inference yields the following measurements for the IGM at $z=0.1$ :

Temperature: $\log(T_0/\text{K}) = 4.45^{+0.08}_{-0.12}$ , corresponding to $T_0 = 28,183^{+5,700}_{-6,804}$ K. This is approximately 7 times hotter than the canonical prediction of $\sim 4000$ K.
Thermal Slope: $\gamma = 1.06^{+0.13}_{-0.09}$ , indicating the low- $z$ IGM is nearly isothermal, in stark contrast to the expected $\gamma \sim 1.6$ .
Ionization Rate: $\log(\Gamma_{\text{HI}}/\text{s}^{-1}) = -13.70^{+0.10}_{-0.08}$ . This value is approximately $4\sigma$ lower than predictions from current UV background synthesis models (e.g., Khaire & Srianand 2019).

These results are consistent across all four redshift bins ( $z=0.1$ to $0.4$), showing a persistently hot and nearly isothermal IGM with a photoionization rate lower than standard models predict.

Alternative Explanations and Tests
To investigate the origin of the broad Ly $\alpha$ lines, the authors tested two alternative hypotheses:

Small-Scale Turbulence: By introducing a turbulent velocity component ( $v_{\text{tur}}$ ) into standard thermal models ( $T_0 \sim 4000$ K), the authors found that a turbulence dispersion of $v_{\text{tur}} \approx 15$ km s $^{-1}$ at $z=0.1$ could reproduce the observed line widths. The required turbulence increases toward lower redshifts, suggesting a continuous source of energy injection.
Instrumental Resolution: The authors tested whether the discrepancy could be caused by an overestimation of the HST COS resolution. By comparing COS spectra with higher-resolution HST/STIS E140M spectra of the same objects, they confirmed that the reported COS LSF is accurate. They demonstrated that reproducing the observed $b$ -distribution with standard thermal models would require an effective resolution of $R \sim 7000$ (FWHM $\sim 43.6$ km s $^{-1}$ ), which contradicts the known instrument performance.

Significance and Claims
The paper claims that the low-redshift IGM is significantly hotter and more isothermal than predicted by standard cosmological models that rely solely on photoionization heating and adiabatic cooling. The authors assert that this discrepancy cannot be explained by instrumental resolution errors. Consequently, the findings imply the existence of either:

New Heating Mechanisms: Processes such as dark matter annihilation, dark photon heating, gamma-ray blazars, or dust heating that become significant around $z \sim 1$ and persist to the present day.
Unresolved Turbulence: Small-scale turbulent motions driven by galaxy formation feedback or other processes not captured in current simulations.

The study emphasizes that the combination of high temperatures and low photoionization rates challenges the current understanding of the IGM's thermal history over the last 10 billion years of cosmic time. The authors conclude that resolving this discrepancy requires either identifying new physical heating sources or incorporating unresolved turbulence into IGM models.