MeerKAT Galaxy Cluster Legacy Survey

• MGCLS is a comprehensive radio survey of 115 galaxy clusters utilizing MeerKAT’s L-band for high-fidelity continuum and polarimetric imaging.
• It employs advanced imaging techniques and machine learning methods to extract, classify, and analyze diffuse radio emissions such as halos, relics, mini-halos, and phoenixes.
• The survey sets new benchmarks in sensitivity and resolution, enabling unprecedented studies of ICM physics, galaxy evolution, and dark matter constraints.

The MeerKAT Galaxy Cluster Legacy Survey (MGCLS) is a comprehensive radio survey targeting galaxy clusters, designed to probe a range of astrophysical phenomena in the intracluster medium (ICM), galaxy evolution, and the cosmic web. MGCLS utilizes the MeerKAT interferometer primarily in the L-band (900–1670 MHz), delivering high-fidelity, deep continuum and polarimetric imaging for 115 galaxy clusters, accompanied by raw visibilities and multi-resolution, wide-band image cubes. The survey's published data releases have established a new benchmark in sensitivity, angular resolution, and field coverage for cluster-scale radio astronomy, enabling detailed studies of non-thermal emission, star formation, magnetic field structure, dark matter signatures, and machine-learning–driven source characterization.

1. Survey Design and Observational Capabilities

The MGCLS comprises 115 cluster fields spanning –80° < Dec < +15° and 0.05 ≲ z ≲ 0.65, with each observed for 6–12 hours (Knowles et al., 2021). The array configuration (up to 64 MeerKAT dishes) yields an angular resolution of ≃8″ (robust=-1.5) and is sensitive to structures up to ≈10′, with typical rms noise of 3–5 μJy beam⁻¹ (full resolution), and 6–15 μJy beam⁻¹ at 15″. Large bandwidth and full-Stokes capability underpin both continuum and polarization studies, HI mapping, and robust in-band spectral index, Faraday, and rotation measure (RM) mapping (Knowles et al., 2021, Kolokythas et al., 5 Sep 2025).

Data products include:

• Raw visibilities for reproducible reprocessing.
• Basic image cubes covering ≈2°×2° (≈4 deg²), including reference-frequency brightness, in-band spectral index, and sub-band images.
• Enhanced products (corrected for the primary beam) for the central 1.2°×1.2°, with five-plane cubes incorporating brightness, spectral index, and their uncertainties.
• Stokes Q/U/V cubes for a subset, enabling RM synthesis and accurate magnetic field diagnostics.
• **Publicly-released, primary-beam–corrected compact source catalogues (~626,000 sources at 1.28 GHz, SNR ≥ 5) with astrometric precision and ~6% flux-scale accuracy (Knowles et al., 2021).

The instantaneous uv-coverage—enabled by the dense core and long baselines—simultaneously recovers compact and diffuse emission, establishing MGCLS as the deepest and most spatially complete 1.28 GHz survey of clusters in the Southern hemisphere (Kolokythas et al., 5 Sep 2025).

2. Diffuse Cluster Radio Emission: Halos, Relics, Mini-halos, and Phoenixes

The MGCLS catalogue of diffuse sources documents 103 confirmed or candidate diffuse radio emissions in 62 of 115 clusters, subdivided as 53% relics (55/103), 32% halos (33/103), 10% mini-halos (11/103), and 3% phoenixes (3/103) (Kolokythas et al., 5 Sep 2025). Source sizes span 55 kpc to >2 Mpc (largest: NW relic in Abell 3667), with rest-frame 1.28 GHz radio powers $P_{1.28\mathrm{GHz}}$ from $10^{22}$ to $>10^{25}$ W Hz⁻¹, and in-band spectral index $\alpha_{908}^{1656}$ down to –3.5 for revived fossil plasma.

Classification criteria:

• Halos: Unpolarized, Mpc-scale, central with respect to X-ray emission, steep spectrum ($\alpha\lesssim -1$), associated with mergers.
• Mini-halos: ≤500 kpc, surrounding a BCG in cool-cores, steep spectrum, often confined by cold fronts.
• Relics: Elongated, peripheral, polarized, tracing merger shocks.
• Phoenixes: Revived, fossil AGN plasma with ultra-steep, curved spectra.

MGCLS pushed the halo radio power–mass relation $P_{1.4\mathrm{GHz}}$–$M_{500}$ to lower masses/powers, with mini-halo detections extending the $P_{1.4}$–$M_{500}$ correlation down by ~0.5 dex. Approximately 58% (60/103) of these diffuse sources are new discoveries (Kolokythas et al., 5 Sep 2025).

3. Methodologies: Imaging, Source Extraction, and Cluster Membership

Imaging and calibration workflows employ direction-independent and (for follow-up) direction-dependent strategies, with multi-scale CLEAN for deconvolution, mosaicking, and robust primary-beam correction (Kolokythas et al., 5 Sep 2025, Knowles et al., 2021). Source extraction is performed with PyBDSF at a 5σ threshold, and catalogue construction includes angular-size cuts and deblending around bright neighbours.

Cluster membership is assigned through probabilistic photo-z estimates from DECaLS g, r, z and WISE W1–W2, leveraging template-fitting with zCluster and calibrating against $\sim 2\times 10^5$ SDSS spectroscopic redshifts (biweight scatter $\sigma_{bw} \approx 0.03$ in $\Delta z/(1+z)$). AGN are removed via X-ray cross-matching, mid-IR color cuts, and upper limits in $L_{1.4\,\mathrm{GHz}}$ (Kesebonye et al., 2022).

Star formation rates (SFRs) are derived from radio luminosities using the Bell (2003) calibration (Chabrier IMF), applying separate formulae above and below the $L_c = 6.4\times 10^{21}$ W Hz⁻¹ threshold.

4. Key Scientific Results: Star Formation, Environmental Effects, and ICM Physics

MGCLS established the first dust-unbiased, statistical census of SFR as a function of environment in intermediate-redshift, massive clusters (Kesebonye et al., 2022). Results include:

• Declining $f_{\mathrm{SF}}(r)$: Fraction of star-forming galaxies declines from $\sim0.23$ at $2 R_{200}$ to $\sim0.12$ at cluster centers.
• Mergers drive elevated SF: Unrelaxed clusters with radio halos/relics exhibit $\sim23$% higher $f_{\mathrm{SF}}$ within $R_{200}$ and a $3\sigma$ higher mass-normalized SFR ($$26.1 \pm 1.4\,M_\odot\,\mathrm{yr}^{-1}/10^{14}\,M_\odot$$ vs $21.5 \pm 1.9$ in non-radio clusters).
• LIRG activity decline: A $\sim4\times$ drop in mass-normalized SFR ($>10\,M_\odot\,\mathrm{yr}^{-1}$) over 2 Gyr agrees with rapid redshift evolution inferred from IR studies.

Interpretation: Environmental quenching progressively suppresses SF toward dense cluster environments, but merger activity (manifested by halos/relics) can temporarily boost SF, especially through the compression and shock processing of the ICM. Rapid declines in luminous SF with decreasing redshift highlight the dominant role of mass accretion history in galaxy growth (Kesebonye et al., 2022).

Abell 3667 observations with MGCLS revealed filamentary detail (down to 3 kpc) in relics, high fractional polarization (p ≈ 70% at edges), and an elongated turbulent radio halo trailing the bullet-like subcluster. Rotation Measure synthesis and polarimetry provided new constraints on magnetic field geometries and draping effects, illustrating the role of merger-driven shocks and turbulence (Gasperin et al., 2021).

5. Non-thermal Phenomena: Dark Matter Constraints and High-z Radio Sources

MGCLS datasets enabled radio-frequency indirect searches for WIMP dark matter annihilation, including generic channels (e.g., $\chi\chi\to b\overline{b}$) and specific particle models (2HDM+S). Analyses integrated observed flux-density upper limits with spatially-resolved synchrotron emission modeling, accounting for uncertainties in DM profile, magnetic fields, and electron diffusion.

• 95% CL limits: For $\chi\chi\to b\overline{b}$, $\langle\sigma v\rangle>10^{-24}$ cm³/s at $m_\chi\approx10$ GeV; thermal relic cross-section reached only at $m_\chi\lesssim1$–3 GeV for optimistic $B_0=10 \mu$G (Lavis et al., 2023, Lavis et al., 2022).
• Complementarity: Radio bounds are competitive with Fermi-LAT for $m_\chi\lesssim5$ GeV. Within 2HDM+S, MGCLS excludes mediators $m_S\approx10$–30 GeV for $m_\chi=5$–20 GeV.

Searches for high-z radio galaxy (HzRG) candidates used MGCLS's sub-0.1 mJy sensitivity, wide bandwidth, and moderate angular resolution, utilizing optical/IR dropout and ultra-steep spectrum (USS, $\alpha<-1.0$) selection. In Abell 2751, 274 HzRG candidates were identified, 179 with measured USS (Knowles et al., 2021).

6. Data-Driven and AI-Based Source Characterization

MGCLS has catalyzed the adoption of machine learning and self-supervised pipelines:

• Self-Supervised Learning (SSL): Models (ResNet50/ViT) trained on ~20,000 crops from MGCLS continuum images (contrast stretching normalization, randomized partitioning, radio-specific augmentations) yield embeddings suitable for compact source count regression (test MSE ≈ 40–50), FRI/FRII classification (F1 ≈ 0.9), and fine-tuning for multi-class morphology (accuracy ≈ 80%). These models match or surpass state-of-the-art astronomy-trained and general-purpose vision backbones (Lastufka et al., 2024).
• Unsupervised Anomaly Detection: VQ-VAEs, trained on unlabeled source cutouts, identify exotic or anomalous morphologies with high efficiency (mean F1 = 0.79, F2 = 0.90, AUC = 0.87 in balanced sets), reducing human screening workload by ≳80% for exotic source fractions ~2.8% (Ventura et al., 20 Jan 2026).
• Faraday RM Deconvolution: Semi-supervised deep U-Nets outperform RMCLEAN in recovering complex or high-RM Faraday spectra from MGCLS L-band polarimetric cubes, delivering a ≈20× speedup, super-resolved φ-space recovery, and detailed magnetic-field mapping in relics, halos, and AGN (Gustafsson et al., 2024).

7. Scientific Outcomes and Legacy

MGCLS has:

• Extended the census and phenomenology of cluster-scale non-thermal phenomena to lower mass, lower power, and high-redshift systems, quantifying scaling relations for halos and mini-halos.
• Provided radio-based evidence for the interplay between ICM shocks, magnetic fields, and non-thermal particle populations, as exemplified by the detailed mapping of surficial brightness edges in radio halos and their one-to-one spatial coincidence with X-ray discontinuities (Botteon et al., 2023).
• Demonstrated the utility and necessity of radio data for star-formation, AGN, and DM studies in dusty environments inaccessible to IR/optical diagnostics.
• Set a contemporary standard for machine-learned, automated, and scalable analysis pipelines—applicable to impending SKA surveys—covering source classification, anomaly discovery, and physical parameter extraction.
• Enabled multi-wavelength and multi-physics synergies, offering commensal HI mapping, polarization diagnostics, and robust cross-matching with optical/X-ray catalogues.

MGCLS thus represents a legacy dataset for cluster radio astrophysics, non-thermal ICM physics, cosmic magnetism, and data-intensive methodologies, providing a template for future SKA-scale cosmological and astrophysical surveys (Knowles et al., 2021, Kolokythas et al., 5 Sep 2025).