🔭 astrophysics

Revealing cosmological fluctuations in 21cm intensity maps with MeerKLASS: from maps to power spectra

This paper outlines the robust analysis framework and early results of the MeerKLASS survey, demonstrating successful detection of cosmological fluctuations in 21cm intensity maps through foreground cleaning and cross-correlation with galaxy surveys, while forecasting the potential of the 64-dish MeerKAT array for wide-sky spectroscopic mapping.

Original authors: Steven Cunnington, Matilde Barberi-Squarotti, José Luis Bernal, Stefano Camera, Isabella P. Carucci, Zhaoting Chen, José Fonseca, Mario Santos, Marta Spinelli, Jingying Wang, Laura Wolz

Published 2026-02-03

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Steven Cunnington, Matilde Barberi-Squarotti, José Luis Bernal, Stefano Camera, Isabella P. Carucci, Zhaoting Chen, José Fonseca, Mario Santos, Marta Spinelli, Jingying Wang, 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 trying to hear a single, faint whisper in a room filled with the roar of a jet engine, the hum of a refrigerator, and the chatter of a crowded party. That is essentially what astronomers are trying to do when they study the universe using 21cm intensity mapping.

This paper describes the work of the MeerKLASS team, a group of scientists using the MeerKAT radio telescope in South Africa to listen to that whisper. Here is a breakdown of their journey, the challenges they faced, and what they found, explained in everyday terms.

The Big Picture: Listening to the "Ghost" of the Universe

The universe is filled with invisible clouds of neutral hydrogen gas. These clouds are the building blocks of galaxies. When these clouds emit a specific type of radio signal (the "21cm line"), it's like a cosmic fingerprint.

The Goal: Instead of trying to see individual galaxies (like taking a photo of every person in a stadium), the team wants to map the glow of the gas itself. This is like trying to map the crowd in a stadium by measuring the collective heat or sound rather than counting individual people.
The Tool: They are using MeerKAT, a massive array of 64 radio dishes. Usually, these dishes work together like a giant microscope (an interferometer) to see fine details. But for this job, the team had to use them differently: as 64 individual "ears" listening to the sky at once. This is called "single-dish mode."

The Challenge: The "Static" Problem

The problem is that the hydrogen signal is incredibly faint. The radio sky is dominated by "foregrounds"—bright, noisy signals from our own galaxy (like static on an old TV) and human-made interference (like Wi-Fi or cell phones).

The Analogy: Imagine trying to hear a baby crying in a room where a rock concert is playing. The baby's cry is the cosmological signal; the rock concert is the foreground noise.
The Solution (Cleaning): The team developed a sophisticated "noise-canceling" system. They use a mathematical technique called Principal Component Analysis (PCA). Think of this as a smart filter that learns what the "rock concert" sounds like (because it's smooth and predictable) and subtracts it out, leaving behind the "baby crying" (the hydrogen signal).
The Catch: Sometimes, the filter is so good it accidentally mutes the baby's cry a little bit too. The team had to create a "transfer function"—a correction tool—to figure out exactly how much of the signal they accidentally removed and add it back in mathematically.

The Innovation: Turning a Sphere into a Cube

Radio telescopes look at the sky in a curved dome (like a bowl). However, to do the complex math needed to find patterns in the universe, scientists prefer to work with a flat, 3D grid (like a cube).

The Problem: Trying to flatten a curved map onto a cube without distorting the shapes is like trying to peel an orange and lay the skin flat on a table without tearing it. If you do it wrong, you stretch or squash the data, leading to wrong conclusions.
The Fix: The team developed a new method to "regrid" the data. They take the curved sky map and redistribute the information into a perfect 3D cube, ensuring the distances between points remain accurate. This allows them to use powerful computer algorithms (like Fast Fourier Transforms) to find patterns.

The Results: Finding the Patterns

Once they cleaned the noise and straightened the map, the team looked for clustering. In the universe, matter isn't scattered randomly; it forms a "cosmic web" with clumps and voids.

The Cross-Check: To prove they actually found the hydrogen signal and not just random noise, they compared their radio map with optical maps of galaxies (like the WiggleZ and GAMA surveys).
The Discovery: They found a strong match. The patterns in the hydrogen gas lined up perfectly with the patterns in the galaxies. This was a 7.7-sigma detection (in the first pilot) and 4-sigma in deeper surveys. In science, this is like flipping a coin and getting heads 20 times in a row—it's a definitive proof that the signal is real.
The Significance: This is the first time a multi-dish array (MeerKAT) has successfully used single-dish mode to map the universe this way. It proves the technique works and paves the way for the future Square Kilometre Array (SKA).

What's Next? The "Wide and Shallow" Strategy

The team is now planning a massive survey covering 10,000 square degrees of the sky (about 1/4 of the entire sky).

The Strategy: Instead of staring deeply at a tiny patch of sky, they will scan a huge area quickly.
Why? The biggest source of error in these maps isn't the telescope's noise; it's "cosmic variance" (the fact that we only have one universe to look at). By looking at a wider area, they get more independent samples of the universe, which reduces the error bars significantly.
The Payoff: This massive survey will help them measure how fast the universe is expanding, test if Einstein's theory of gravity holds up over vast distances, and even look for clues about the very first moments after the Big Bang (primordial non-Gaussianity).

Summary

The MeerKLASS team has successfully turned a 64-dish radio telescope into a giant "cosmic microphone." They built a new pipeline to filter out the universe's static, straightened the curved sky into a usable 3D map, and proved that they can detect the faint whisper of hydrogen gas across the cosmos. This success lays the groundwork for the next generation of radio telescopes to map the entire history of the universe.

Technical Summary: Revealing cosmological fluctuations in 21cm intensity maps with MeerKLASS

Problem Statement
Mapping the large-scale structure of the Universe via the redshifted 21cm emission line of neutral hydrogen (Hi) offers a powerful, efficient method for charting matter fluctuations across vast cosmological volumes. However, prior to the MeerKAT array, this technique was limited to pioneering efforts using single-dish telescopes (e.g., GBT, Parkes) with restricted sky coverage and sensitivity. A significant challenge arises when applying this technique to the MeerKAT array, which is primarily designed as an interferometer. The array's baseline distribution is ill-suited for the very large angular scales required for cosmological intensity mapping. Furthermore, extracting the faint cosmological Hi signal is complicated by dominant astrophysical foregrounds (synchrotron and free-free emission) that are orders of magnitude brighter, radio frequency interference (RFI), instrumental systematics, and the inherent faintness of the Hi brightness temperature. Validating the "single-dish" mode (auto-correlation of voltage signals from individual dishes) for a multi-dish array and developing a pipeline to produce unbiased measurements has presented major technical hurdles.

Methodology
The paper outlines the robust analysis framework developed by the MeerKLASS collaboration to overcome these challenges, moving from calibrated maps to power spectra.

Observational Strategy: MeerKLASS utilizes the 64-dish MeerKAT array in single-dish auto-correlation mode. A fast scanning strategy (back-and-forth azimuth scans at constant elevation) is employed to suppress slow instrumental gain variations. Observations have transitioned from the L-band (limited redshift coverage due to RFI) to the UHF-band, offering a wider redshift window ( $0.4 < z < 1.45$ ) and a significantly larger comoving volume.
Foreground Cleaning: To remove spectrally smooth astrophysical foregrounds, the collaboration employs "blind" component separation techniques, primarily Principal Component Analysis (PCA). The paper details the development of a multiscale extension (mPCA) using wavelet transforms to separate large- and small-scale contaminants, improving signal preservation.
Signal Loss Correction: Blind cleaning inevitably removes a fraction of the cosmological signal. The authors utilize a foreground transfer function, $T(k)$ , constructed by injecting mock Hi signals into real contaminated data, applying the cleaning pipeline, and measuring the fractional suppression of the power spectrum. This empirical approach corrects for signal loss and accounts for complex signal-foreground couplings. Recent theoretical work (Chen 2025) is cited to clarify that the transfer function acts as a renormalization of the estimator window function, independent of the cosmological model.
Regridding: To enable Fourier-based power spectrum estimation, spherical sky maps are regridded into a 3D Cartesian comoving volume. The paper addresses systematic biases introduced by this process, specifically the damping of small-scale power due to nearest-grid-point (NGP) mass assignment and aliasing. Corrections involving the Fourier transform of the mass assignment window function are applied to recover accurate power spectra.
Covariance Estimation: Moving beyond analytical Gaussian error estimates, the authors utilize the scatter of the transfer function across mock injections to estimate the full covariance matrix. This method captures contributions from sample variance, residual systematics, and the variance introduced by signal loss projection.

Key Contributions and Results

Robust Analysis Pipeline: The paper presents a comprehensive pipeline that successfully validates single-dish intensity mapping on a multi-dish array, addressing calibration, map-making, foreground cleaning, signal loss correction, and regridding.
Cross-Correlation Detections: The collaboration reports statistically significant detections of cosmological clustering by cross-correlating MeerKLASS intensity maps with overlapping optical galaxy surveys:
- WiggleZ: A $7.7\sigma$ detection using early L-band pilot data ( $\sim200$ deg $^2$ ).
- GAMA: A $>4\sigma$ detection using deep-field L-band data ($236$ deg $^2$ ).
- Re-analysis: A refined analysis of the WiggleZ data using mPCA confirmed the detection at $\sim6\sigma$ , demonstrating improved foreground cleaning and reduced signal loss.
- These measurements constrain the combined parameter $\Omega_{HI}b_{HI}r$ to approximately $0.86-0.93 \times 10^{-3}$ .
Auto-Power Spectrum Limits: The deep-field L-band data has reached a regime where cosmological Hi fluctuations begin to dominate thermal noise on large scales. The paper presents the tightest upper limits to date on the Hi auto-power spectrum, with data closely following the expected signal-plus-noise model.
Stacking Analysis: The paper demonstrates the detection of Hi emission via spectral line stacking onto GAMA galaxy positions ( $7.45\sigma$ ), providing a complementary method to power spectrum analysis and a tool for validating calibration pipelines.
Forecasts: Using Fisher matrix forecasts for the planned 10,000 deg $^2$ UHF-band survey (2,500 hours), the paper projects that MeerKLASS will deliver competitive constraints on the growth rate of structure ( $f(z)$ ) and primordial non-Gaussianity ( $f_{NL}$ ), particularly when combined with multi-tracer approaches using the Rubin Observatory.

Significance
The paper claims that these results represent a key milestone: the first robust demonstration of single-dish intensity mapping using a multi-dish interferometric array. By overcoming the systematic challenges inherent to this mode of observation, MeerKLASS has established a viable pathway for precision radio cosmology in the pre-SKA era. The work lays the groundwork for full-scale cosmological analyses with the extended MeerKLASS survey and the future Square Kilometre Array (SKAO). The successful development of the analysis framework, including the handling of signal loss and the construction of unbiased transfer functions, provides a critical foundation for extracting cosmological information from intensity maps, enabling independent tests of gravity and inflationary physics that complement optical surveys.