🔭 astrophysics

A direct detection of neutral hydrogen intensity mapping on Mpc scales at $z\approx 0.32$ and $z\approx 0.44$

Using MeerKAT radio telescope observations, this study reports the first direct detection of the neutral hydrogen intensity mapping power spectrum on Mpc scales at redshifts 0.32 and 0.44, achieving statistical significances up to 9.18 $\sigma$ through advanced foreground mitigation techniques.

Original authors: Sourabh Paul, Zhaoting Chen, Mario G. Santos, Laura Wolz

Published 2026-07-07

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Sourabh Paul, Zhaoting Chen, Mario G. Santos, Laura Wolz

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

Imagine the universe is a giant, dark ocean. In this ocean, there are invisible islands made of neutral hydrogen gas. These islands are the building blocks of galaxies, but they are too faint to see directly with our eyes or even most telescopes. They whisper a secret signal: a specific radio "hum" at a frequency of 21 centimeters.

This paper is about a team of astronomers who finally managed to "hear" this whisper across vast distances, not by listening to individual islands, but by measuring the collective hum of the entire ocean.

Here is a breakdown of their journey, using simple analogies:

1. The Challenge: Listening in a Storm

The team used the MeerKAT telescope in South Africa, which is like a massive, high-tech ear made of 64 dishes. Their goal was to map the distribution of this hydrogen gas at two specific times in the universe's history (when the universe was about 9 and 10 billion years old).

However, trying to hear the faint cosmic hum is like trying to hear a single cricket chirping in the middle of a rock concert.

The Concert (Foregrounds): The sky is filled with bright, loud radio sources (like our own Milky Way galaxy and distant radio galaxies) that drown out the faint hydrogen signal.
The Static (RFI): There is also "Radio Frequency Interference" (RFI)—man-made noise from satellites, cell towers, and even lightning. This is like static on a radio that sounds exactly like the signal you are looking for.

2. The Strategy: The "Silent Room" Technique

Instead of trying to turn down the volume of the loud concert (which is nearly impossible), the team used a clever trick called Foreground Avoidance.

Imagine the loud concert noise is only allowed to sit in the front rows of a theater, while the faint cricket chirps are in the back. The team decided to only listen to the "back rows" of the data. They mathematically filtered out the parts of the signal where the loud noise lives, leaving them with a "quiet room" where they could hopefully hear the hydrogen.

3. The Problem: The Ghost in the Machine

Even after clearing out the loud noise, they found something strange. There were faint, ghostly ripples in their data that looked like the signal but weren't. These were low-level RFI—tiny bits of interference that slipped through the cracks.

To fix this, they developed two different "noise-cancellation" strategies:

Strategy A (The "Bad Seat" Approach): They looked at the raw data from every pair of telescope dishes. If a specific pair of dishes was acting up and picking up weird noise, they threw out all the data from that pair for that time. It's like saying, "If this specific seat in the theater is squeaking, we won't use any data from that seat."
Strategy B (The "Outlier" Approach): They looked at the final processed data. If a specific data point was wildly different from what pure random noise should look like (like a sudden spike in volume), they flagged and removed just that specific point. This is like saying, "We'll keep the data from the squeaky seat, but we'll ignore that one loud squeak."

4. The Result: Hearing the Universe

By using these methods, they successfully detected the Power Spectrum of the hydrogen.

What is a Power Spectrum? Imagine you are looking at a crowd of people. You can't see everyone clearly, but you can measure how "clumped" they are. Are they standing in tight groups, or are they spread out evenly? The power spectrum tells the astronomers how the hydrogen gas is clumped together across the universe.
The Success: They found a clear signal.
- Using the "Bad Seat" method, they were 3.2 to 3.5 times more confident that the signal was real than random noise.
- Using the "Outlier" method, they were 5.9 to 9.2 times more confident. In the world of science, a "9.2 sigma" result is a massive, undeniable discovery.

5. What They Learned About the "Islands"

Once they confirmed the signal was real, they used it to guess the properties of the hydrogen gas:

The "Wiggle" Factor: They measured how fast the gas clouds are moving and wobbling inside galaxies (called "velocity dispersion"). It's like measuring how much a crowd is shuffling around.
The "Noise" Level: They estimated the "shot noise," which is the statistical fuzziness caused by the fact that galaxies are discrete objects, not a smooth fluid.

The Big Picture

This paper is a milestone because it is the first time anyone has directly detected this specific type of hydrogen signal using an interferometer (a telescope array) without needing to cross-reference it with a map of visible galaxies.

Think of it this way: Before this, astronomers could only guess where the hydrogen was by looking at the bright stars (galaxies) that live inside it. Now, they have learned to listen to the gas itself, even when the stars are too faint to see. They proved that with the right tools and noise-canceling techniques, we can map the invisible skeleton of the universe.

In short: The team built a super-sensitive radio ear, figured out how to ignore the loud cosmic concert and the static interference, and successfully heard the faint, collective hum of the universe's hydrogen gas for the first time.

Technical Summary: Direct Detection of Neutral Hydrogen Intensity Mapping on Mpc Scales

Problem and Context
The 21-cm hyperfine transition of neutral hydrogen (HI) serves as a primary tracer for the matter distribution of the Universe. However, the inherent weakness of the 21-cm signal has historically restricted HI surveys to the nearby Universe ( $z \lesssim 0.1$ ) or required cross-correlation with galaxy surveys to overcome calibration challenges and foreground contamination. While intensity mapping (IM) offers a pathway to probe the three-dimensional matter distribution at higher redshifts by averaging collective emission from unresolved galaxies, direct detection of the HI auto-power spectrum without cross-correlation has remained elusive due to the overwhelming presence of bright radio foregrounds and systematics.

Methodology
This study reports the first direct detection of the HI intensity auto-power spectrum using interferometric observations from the MeerKAT radio telescope in South Africa. The analysis utilized approximately 96 hours of data collected in 2018 from the DEEP2 field ( $\alpha = 04^h13^m26.4^s, \delta = -80^\circ$ ), focusing on two redshift bins: $z \approx 0.32$ (1077.5 MHz) and $z \approx 0.44$ (986 MHz).

The methodology employed a "delay spectrum" approach, performing a Fourier transform along the frequency axis of gridded visibilities to separate the smooth-spectrum foregrounds from the cosmological signal. Key technical components included:

Data Processing: Calibration was performed using the processMeerKAT pipeline, followed by phase-only self-calibration.
Foreground Avoidance: The analysis restricted the power spectrum estimation to modes above a conservative foreground wedge defined by $k_\parallel = 0.3 k_\perp$ .
Systematic Mitigation: Two distinct frameworks were developed to address residual low-level broadband Radio Frequency Interference (RFI) and systematics:
1. Baseline Flagging (BLF): A conservative approach where antenna pairs exhibiting peaks in the delay power spectrum (above $5\sigma$ of thermal noise) were flagged per scan before power spectrum calculation.
2. UV-Delay Flagging (UVDF): A method applied after gridding, where specific $k$ -pixels in the 3D delay space exceeding $5\sigma$ of the expected thermal noise were flagged. This approach retains more data than BLF by only removing contaminated modes rather than entire baselines.
Validation: The pipeline was validated through jackknife tests (excluding observation blocks), null tests (cross-correlating different redshift bins to check for frequency-coherent systematics), and simulations of foreground scatter due to calibration errors.

Key Results
The study achieved independent measurements of the HI power spectrum at $z \approx 0.32$ and $z \approx 0.44$ :

Statistical Significance: The BLF method yielded detections at $3.2\sigma$ ( $z=0.32$ ) and $3.5\sigma$ ( $z=0.44$ ). The UVDF method, which mitigated systematics more effectively while retaining data, enhanced the significance to $5.9\sigma$ and $9.18\sigma$ respectively (after excluding the first two $k$ -bins most susceptible to systematics).
Signal Amplitude: The measured power spectrum amplitude is approximately $\sim 1 \, \text{mK}^2 \text{Mpc}^3$ , decreasing to $\sim 0.1 \, \text{mK}^2 \text{Mpc}^3$ at smaller scales. This is consistent with expected shot noise levels and attenuation from Finger-of-God (FoG) effects.
Systematics Characterization: The analysis identified weak, broadband RFI concentrated near the $u=0$ line in the $uv$-plane. Null tests confirmed that the chosen observation window and flagging strategies successfully isolated the cosmological signal from these systematics, as no significant correlation was found between the two redshift bins.
Astrophysical Constraints: Using a halo model formalism, the authors constrained the HI shot noise and velocity dispersion ( $\sigma_v$ ). While parameter degeneracies prevented full constraints on the HI mass function (HIMF) without priors, imposing constraints on the HI density ( $\Omega_{HI}$ ) from previous stacking and DLA studies allowed for preliminary constraints on the velocity dispersion ( $\sigma_v \approx 54^{+50}_{-35} \, \text{km/s}$ at $z \approx 0.44$ ) and shot noise.

Significance and Claims
The authors claim this work represents a significant step toward precision cosmology with HI intensity mapping using next-generation radio telescopes. The primary contribution is the demonstration that the HI auto-power spectrum can be directly detected on Mpc scales at $z \sim 0.3-0.4$ without relying on cross-correlations with galaxy surveys.

The paper emphasizes that while the "foreground avoidance" technique is robust, the presence of weak, broadband RFI poses a critical risk to sensitive cosmological observations. The development and validation of the UVDF and BLF flagging strategies provide a necessary toolkit for future surveys (such as MIGHTEE and Laduma) to achieve high-precision measurements of the small-scale HI power spectrum, ultimately enabling detailed studies of galaxy evolution and the HI halo model from $z \sim 0.0$ to $z \sim 1.0$ . The results confirm that with sufficient integration time and rigorous systematic control, interferometric IM can probe the nonlinear regime and small scales inaccessible to single-dish surveys.