Tracing Large-scale Structure with the MeerKLASS On-the-Fly Survey: Angular Clustering of Radio Sources at 816 MHz

Technical Summary: Tracing Large-scale Structure with the MeerKLASS On-the-Fly Survey

Problem and Motivation
Radio continuum surveys offer a unique, dust-unbiased window into the large-scale structure (LSS) of the Universe, tracing the matter density field through radio-loud active galactic nuclei (AGN) and star-forming galaxies (SFGs). However, robust angular clustering measurements from radio catalogues are non-trivial due to spatially varying completeness driven by position-dependent noise, complex source morphology (where single physical galaxies are resolved into multiple catalogue components), and the general lack of direct redshift information. Previous measurements using surveys like FIRST and NVSS established the clustering of bright, AGN-dominated populations, but there remains a need for high-fidelity clustering measurements at intermediate $\sim$GHz frequencies using modern interferometers with well-characterized depth variations and overlap with contemporary optical spectroscopic programs. This paper addresses these challenges by presenting the first angular clustering analysis of the MeerKAT Large Area Synoptic Survey (MeerKLASS) UHF on-the-fly (OTF) continuum Data Release 1 (DR1).

Methodology
The analysis utilizes the MeerKLASS UHF DR1 dataset, which covers $\sim$800 deg$^2$ within the DESI footprint at a reference frequency of 816 MHz. The survey employs a constant-elevation OTF strategy with cross-linked rising and setting scans, resulting in a non-uniform depth pattern that must be explicitly modeled.

1. Sample Definition: A conservative clustering sample was constructed from the 95,483 unique sources in the DR1 source catalogue (SRL). The fiducial selection applies:

   • A flux-density threshold of $S_{816} > 2$ mJy.
   • A depth-controlled mask restricting the analysis to regions with local RMS noise $\sigma_{local} \le 250$ $\mu$Jy beam$^{-1}$.
   • A "compact-only" criterion ($S\_Code = 'S'$ and $N\_Gaus = 1$) to mitigate small-scale excess power caused by fragmented multi-component radio galaxies.
   • A conservative Right Ascension (RA) edge cut ($RA \le 184.07^\circ$) to ensure uniformity between data and randoms.
     This resulted in a final clustering catalogue of 39,483 sources.

2. Mask and Randoms: A binary survey mask was constructed from the geometric footprint and the RMS depth map. Random catalogues were generated uniformly within this mask to trace the angular selection function, validated by checking the uniformity of the data-to-random surface density ratio.

3. Correlation Measurement: The angular two-point correlation function, $w(\theta)$, was measured using the Landy–Szalay estimator over 20 logarithmically spaced bins ($0.02^\circ \le \theta \le 10^\circ$). Uncertainties and bin-to-bin covariances were estimated using a spatial jackknife resampling technique with 30 patches. An integral constraint (IC) correction was applied to account for the finite survey footprint.

4. Modeling and Bias Inference:

   • Phenomenological Fit: The signal was characterized using a power-law model $w(\theta) \approx A(\theta/1^\circ)^{1-\gamma}$ with a fixed slope $\gamma=1.8$.
   • Physical Interpretation: The measured clustering was interpreted by fitting $\Lambda$CDM projected-matter templates ($w_{DM}(\theta)$) computed with CAMB and the Limber approximation. Since individual redshifts are unavailable, the analysis bracketed the dominant systematic uncertainty (the redshift distribution $N(z)$) using two priors derived from T-RECS semi-empirical simulations: an AGN-dominated prior and a TOTAL prior (AGN + SFG). An effective large-scale bias ($b_{eff}$) was inferred by scaling the template to the data.

Key Results

• Detection: A positive clustering signal was detected over the range $0.02^\circ \lesssim \theta \lesssim 10^\circ$. The measurement is stable to variations in the depth mask and flux threshold on intermediate and large scales.
• Power-Law Amplitude: Fitting the intermediate-scale signal ($0.112^\circ \le \theta \le 1.36^\circ$) with a fixed slope $\gamma=1.8$ yields an amplitude of $A(1^\circ) = (1.434 \pm 0.475) \times 10^{-3}$, corresponding to $\log_{10} A = -2.843^{+0.124}_{-0.175}$.
• Effective Bias: The inferred effective bias depends significantly on the assumed redshift distribution:
  • AGN prior: $b_{eff} = 1.998 \pm 0.350$.
  • TOTAL prior: $b_{eff} = 1.530 \pm 0.265$.
    The spread between these values ($\Delta b_{eff} \approx 0.47$) is identified as the dominant modelling uncertainty, arising from the unknown population mix of the flux-limited sample.
• Correlation Length: Limber-inverting the power-law amplitude yields spatial correlation lengths of $r_0 = 6.18 \pm 1.13$ $h^{-1}$Mpc (AGN prior) and $r_0 = 5.59 \pm 1.02$ $h^{-1}$Mpc (TOTAL prior).
• Systematics: Relaxing the compactness criterion resulted in a pronounced upturn in $w(\theta)$ at small scales ($\theta < 0.1^\circ$), confirming the necessity of the compact-only selection for robust large-scale clustering measurements.

Significance and Claims
The paper establishes MeerKLASS UHF DR1 as a new, wide-area, intermediate-frequency dataset for radio-continuum clustering. The authors claim that their results demonstrate the viability of OTF interferometric continuum imaging for precision large-scale-structure measurements, bridging the gap between classic all-sky GHz surveys (e.g., NVSS) and new-generation low-frequency surveys (e.g., LoTSS).

The study highlights that while current results are limited by the uncertainty in the redshift distribution ($N(z)$), the overlap with the DESI footprint provides a clear path forward. The authors state that future releases, incorporating overlapping optical/IR spectroscopy for improved redshift calibration, will enable population-split clustering and bias evolution measurements with substantially reduced modelling uncertainty. The current work serves as a foundational step, providing a conservative, well-controlled measurement that validates the survey strategy and data products for cosmological applications.