Assessing the large-scale angular clustering of UNIONS Lyman Break Galaxies via cross-correlations

Technical Summary: Assessing the Large-Scale Angular Clustering of UNIONS Lyman Break Galaxies via Cross-Correlations

Problem Statement
Lyman-break galaxies (LBGs) at redshifts $z > 2$ are potent tracers of the high-redshift Universe, offering unique constraints on the growth of structure and primordial non-Gaussianity ($f_{\text{loc}}^{\text{NL}}$). However, measuring their large-scale angular clustering using wide-field photometric surveys like the Ultraviolet Near Infrared Optical Northern Survey (UNIONS) is severely hampered by spatially varying imaging systematics. Variations in Point Spread Function (PSF) depth, seeing, and Galactic extinction across the survey footprint introduce artificial fluctuations in the observed galaxy density. These systematics are particularly problematic for faint, magnitude-limited samples near the survey depth, where small changes in photometric conditions significantly affect source selection completeness. Previous studies have shown that even linear or non-linear mitigation techniques often fail to fully remove these contaminants, rendering the LBG auto-angular power spectrum unreliable for cosmological inference on large scales.

Methodology
This study evaluates the feasibility of using UNIONS-selected LBGs as cosmological probes by shifting focus from auto-correlations to cross-correlations with external tracers less sensitive to the same imaging systematics. The methodology proceeds as follows:

1. Data Selection: LBGs are selected from the UNIONS $ugri$ GAaP catalog (covering $\sim 3,580$ deg$^2$) using a $u$-dropout criterion ($23.5 < r < 24.2$) designed to isolate galaxies at $z \sim 2.5$. The sample yields an average surface density of $\sim 1,400$ galaxies deg$^{-2}$.
2. Systematic Characterization: The dependence of the LBG overdensity on survey characteristics (PSF depth, seeing) and astrophysical templates (Galactic extinction, stellar density) is quantified. Strong correlations are observed, particularly with PSF depth and size.
3. Mitigation Strategies: Two correction methods are applied to the LBG density maps:
   • Linear Deprojection: Implemented via the namaster code, modeling the observed density as a linear combination of the true field and systematic templates.
   • Non-linear Regression: Using the regressis code with Random Forest (RF) algorithms to capture complex, non-linear dependencies between tracer density and imaging systematics.
4. Validation with Mocks: A contamination-decontamination pipeline is developed using simulated Gaussian fields. Mock LBG maps are generated, contaminated using weights derived from the actual UNIONS data (via RF), and then corrected. This validates the ability of the methods to recover input signals in the presence of known contamination.
5. Cross-Correlation Analysis: The corrected UNIONS LBG maps are cross-correlated with:
   • Planck PR4 CMB lensing convergence maps.
   • Quasar samples from DESI DR1 and Quaia.
     The angular power spectra ($C_\ell$) are computed using the pseudo-$C_\ell$ estimator with mode-coupling corrections.

Key Results

• Failure of Auto-Correlation: The LBG auto-angular power spectrum remains heavily contaminated by imaging systematics even after mitigation. The measured power at large scales ($\ell \lesssim 100$) exceeds theoretical predictions by orders of magnitude. While mitigation reduces power, it does not recover a spectrum consistent with theoretical models, confirming that the auto-spectrum is unsuitable for cosmological inference in this dataset.
• Robustness of Cross-Correlation: In contrast, the cross-power spectrum between UNIONS LBGs and external tracers (CMB lensing and DESI QSOs) is robust. Systematic contaminants in the LBG map are largely uncorrelated with the CMB lensing or DESI QSO fields.
  • The LBG-CMB lensing cross-spectrum shows a positive correlation consistent with theoretical expectations in amplitude.
  • Residual systematics primarily manifest as an increase in the variance of the estimator at large angular scales rather than introducing a significant bias in the recovered signal.
• Mock Validation: Simulations confirm that while auto-spectra are sensitive to the specific assumptions of the underlying clustering model and the accuracy of the decontamination weights, cross-spectra remain stable and recover the input signal even when the contamination model is imperfect or non-linear.

Significance and Claims
The paper establishes that UNIONS-selected LBGs are reliable tracers for cross-correlation cosmology at $z \sim 2.5$, despite the presence of complex imaging systematics that preclude their use in auto-correlation studies. The primary significance of this work is the demonstration that cross-correlation techniques provide a powerful and robust avenue for extracting cosmological information from photometric high-redshift galaxy samples.

The authors present the first detection of the cross-correlation signal between $u$-dropout LBGs and the Planck CMB lensing map over an unprecedentedly large sky footprint ($\sim 6.6\%$ of the sky). While the current analysis serves as a proof of concept using conservative modeling, it validates the methodology required for future, more precise constraints on cosmological parameters, including primordial non-Gaussianity, using Stage-IV photometric surveys. The study highlights that for faint, high-redshift samples, cross-correlation with external datasets is not merely an alternative but a necessary strategy to bypass the limitations imposed by survey inhomogeneities.