4MOST Cosmology Redshift Survey (CRS): Clustering properties of CRS BG and LRG target catalogues

TL;DR

The paper validates the 4MOST CRS BG and LRG target catalogues (selected from Legacy Survey DR10.1) by examining angular clustering, masking effects, and redshift distributions through cross-correlation with DESI DR1 and clustering-redshift techniques. It demonstrates that appropriate masks (including unWISE W1 masks for LRGs) yield stable $w(\theta)$ and cross-cap consistency, and that Limber scaling across BG magnitude slices collapses to a common relation when using a broken-power-law $\,\xi(r)"$ model and $N(z)$ from DESI DR1. Clustering-redshift estimates corroborate BG $N(z)$ shapes at faint magnitudes, while LRG clustering follows a power law with modest redshift evolution and is well described by HOD modelling, including photo-$z$ smearing. Together, these tests establish uniform clustering statistics and reliable redshift distributions, supporting precise large-scale structure analyses with 4MOST CRS and enabling future cross-correlations with LSST and Euclid. The results provide a robust foundation for early CRS cosmology and for calibrating photometric redshifts in southern-wide imaging surveys.

Abstract

The 4MOST Cosmology Redshift Survey (CRS) will obtain nearly 5.4 million spectroscopic redshifts over $\sim5700$\,deg$^2$ to map large-scale structure and enable measurements of baryon acoustic oscillations (BAOs), growth rates via redshift-space distortions, and cross-correlations with weak-lensing surveys. We validate the target selections, photometry, masking, systematics and redshift distributions of the CRS Bright Galaxy (BG) and Luminous Red Galaxy (LRG) target catalogues selected from DESI Legacy Surveys DR10.1 imaging. We measure the angular two-point correlation function, test masking strategies, and recover redshift distributions via cross-correlation with DESI DR1 spectroscopy. For BG, we adopt Legacy Survey \texttt{MASKBITS} that veto bright stars, SGA large galaxies, and globular clusters; for LRG, we pair these with an unWISE W1 artefact mask. These choices suppress small-scale excess power without imprinting large-scale modes. A Limber-scaling test across BG $r$-band magnitude slices shows that, after applying the scaling, the $w(θ)$ curves collapse to a near-common power law over the fitted angular range, demonstrating photometric uniformity with depth and consistency between the North (NGC) and South (SGC) Galactic Caps. Cross-correlations with DESI spectroscopy recover the expected $N(z)$, with higher shot noise at the brightest magnitudes. For LRGs, angular clustering in photo-$z$ slices ($0.4\le z<1.0$) is mutually consistent between the DECaLS and DES footprints at fixed $z$ and is well described by an approximate power law once photo-$z$ smearing is accounted for; halo-occupation fits yield results consistent with recent LRG studies. Together, these tests indicate that the masks and target selections yield uniform clustering statistics, supporting precision large-scale structure analyses with 4MOST CRS.

Figures (11)

• Figure 1: Top panel: Footprint of CRS-BG and CRS-LRG as well as other surveys. Bottom panel: Footprint of the 4MOST-CRS SGC sky and the DES photometric survey. The regions of 4MOST-CRS outside DES used DECaLS photometric data. This illustrates that the SGC footprint CRS is mostly using DES photometric data.
• Figure 2: Angular correlation functions for different masking choices. Left panels: BG; right panels: LRG. Upper panels: $w(\theta)$ with 36 jackknife regions for uncertainties. Lower panels: differences relative to the adopted final masking, shown as $\Delta w(\theta)/\sigma(\theta)$ with $\Delta w(\theta) \equiv w_{\rm final}(\theta) - w_{\rm mask}(\theta)$ and $\sigma(\theta) \equiv [\sigma^2_{\rm JK,mask}(\theta)+\sigma^2_{\rm JK,final}(\theta)]^{1/2}$. Negative values indicate that the final mask decreases the measured clustering amplitude compared to a given configuration, as expected when stricter masks remove contaminants. The LRG panels show a clear monotonic approach towards zero with stricter masking, consistent with the removal of star-spike and unWISE artefact detections misclassified as LRGs. BG is largely stable but still benefits from robust bright-star masking. Dashed lines in the upper panels show power-law fits (Equation \ref{['eq:wt_power']}) for the final masking used for each tracer.
• Figure 3: Stacked two-dimensional density maps of CRS BG (top row) and LRG (bottom row) targets around their nearest Gaia stars, split into four Gaia magnitude bins. Axes show separations in RA and Dec scaled by the bright-star mask radius $R_{\rm BS}$. The black circle marks the MASKBIT 1 radius (available only for $G<13$); the blue circle marks the MASKBIT 11 radius, twice the MASKBIT 1 radius, applied to all Gaia and Tycho 2 stars with $G<16$. Colours indicate $\log_2(\rho/\bar{\rho})$, where $\rho$ is the per-pixel target density and $\bar{\rho}$ is the mean in the annulus $1.1<R/R_{\rm BS}<7$. The stacks show over-densities around stars with $G>13$ in both tracers, supporting the use of MASKBIT 11. The thin, near-horizontal features aligned with constant Declination are bleed trails, not optical diffraction spikes.
• Figure 4: Examples of CCD saturation bleed trails ("blooming") in the DESI Legacy Imaging Surveys, shown as cut-outs from the Legacy Surveys Sky Viewer around two randomly selected stars with $8<G<12$. The thin, near-horizontal features at approximately constant Declination are bleed trails—electronic artefacts distinct from the $\sim45^\circ$ optical diffraction spikes, which are also visible. Axes show $\Delta\mathrm{RA}/R_{\mathrm{BS}}$ and $\Delta\mathrm{Dec}/R_{\mathrm{BS}}$, where $R_{\mathrm{BS}}$ is the bright–star mask radius. LS Sky Viewer Credit: Legacy Surveys / D. Lang (Perimeter Institute).
• Figure 5: Upper panel: Redshift distributions $N(z)$ of CRS BG-like targets from DESI DR1 in five $r$-band magnitude slices, with BE model fits (solid lines) and comparison to the DESI BGS sample (dashed line). Middle: Projected correlation function $w_p(r_p)$ for the same slices, showing the scale-dependent clustering strength used to fit the spatial correlation parameters for Limber scaling. Error bars for $w_p(r_p)$ are calculated using 36 Jackknife regions. Bottom: Normalised histogram of absolute magnitude for CRS BG-like and DESI BGS. In the lower panel, absolute r-band magnitudes for DESI BGS are calculated using Kcorrect v5blanton_k-corrections_2007.