The cosmological analysis of DES 3$\times$2pt data from the Effective Field Theory of Large-Scale Structure

TL;DR

This work applies the EFTofLSS framework at one-loop order to DES Year 3 3×2pt data, modeling galaxy clustering, galaxy–galaxy lensing, and cosmic shear in angular space with careful scale cuts and theoretical priors. The authors validate their pipeline against simulations (Buzzard) and synthetic data, quantify theory and observational systematics, and publicly release the PyFowl code. They obtain baseline ΛCDM constraints from DES Y3 3×2pt that yield $S_8 = 0.833 \pm 0.032$, $\Omega_m = 0.272 \pm 0.022$, and $h = 0.773 \pm 0.049$, with consistency within ~$2\sigma$ of Planck and BOSS results and a novel indication that the projected angular statistics probe the matter–radiation equality scale to inform $H_0$. The analysis highlights how modeling choices, scale cuts, and priors impact cosmological inferences and shows the potential for EFTofLSS to improve cross-probe consistency and future photometric surveys (e.g., DESI, Euclid, LSST). The work also documents a structured approach to perturbative convergence, theory-error marginalisation, and integration with observational systematics, establishing a robust framework for interpreting projected large-scale structure data. Overall, this EFTofLSS-based 3×2pt pipeline advances precision cosmology with photometric surveys and sets the stage for next-generation analyses.

Abstract

We analyze the Dark Energy Survey (DES) Year 3 data using predictions from the Effective Field Theory of Large-Scale Structure (EFTofLSS). Specifically, we fit three two-point observables (3$\times$2pt), galaxy clustering, galaxy-galaxy lensing, and cosmic shear, using the one-loop expressions for the projected angular correlation functions. We validate our pipeline against numerical simulations and we check for several internal consistencies before applying it to the observational data. Fixing the spectral tilt and the baryons abundance, we measure $S_8=0.833\pm 0.032$, $Ω_m = 0.272\pm 0.022$, and $h = 0.773\pm 0.049$, to about $3.8\%$, $8.1\%$, and $6.3\%$, at $68\%$CL, respectively. Our results are consistent at the $\sim 1.5-2σ$ level with those from Planck and the BOSS full-shape analyses, as well as with those from DES collaboration 3$\times$2pt analysis combined with a Big-Bang Nucleosynthesis prior and a Planck prior on $n_s$. The shift in the posterior compared to DES collaboration results highlights the impact of modeling, scale cuts, and choice of prior. The theory code and likelihood used for our analyses, \texttt{PyFowl}, is made publicly available.

Figures (15)

• Figure 1: Triangle plots of $\Lambda$CDM cosmological parameters from the EFTofLSS analysis of DES Y3 3$\times$2pt data, with $\omega_b$ and $n_{s}$ set to BBN and Planck preferred values, respectively. Left panel: Results using either the MagLim or redMaGiC sample. Right panel: In comparison to the MagLim results are shown constraints from i) DES Y3 3$\times$2pt analysis by the DES collaboration DES:2021wwk where the publicly-released MCMC chain products are post-processed with a BBN prior on $\omega_b$ and a Planck prior on $n_s$; ii) BOSS galaxy clustering power spectrum and bispectrum analysis from the EFTofLSS at one loop DAmico:2022osl; and iii) CMB data from Planck with free neutrino mass Planck:2018vyg.
• Figure 2: DES Y3 redshift distributions of source galaxies (upper panel) and lens galaxies selected with MagLim or redMaGiC (lower panels). Distributions are normalised such their integral over $z$ is one. As explained in the main text, when analysing the MagLim sample only the first four redshift bins are used. The dotted lines are the redshifts determined from the photometric data, whereas the continuous lines together with the shaded regions represent the $1\sigma$-confidence intervals obtained from calibrating the redshifts to spectroscopic data as explained in DES:2020sjz and sec. \ref{['sec:observational']}. Those are used as prior on photo-$z$ uncertainties in the DES collaboration analysis DES:2021wwk and the present analysis. The thin black lines are obtained using the photo-$z$ parameters from the best-fits in our cosmological analysis. The main cosmological results presented in this analysis, when not explicitly specified, are obtained using the first four redshift bins of MagLim sample, while the redMaGiC sample is mainly used for comparison purposes.
• Figure 3: DES Y3 two-point angular correlation functions: galaxy clustering $w$ and galaxy-galaxy lensing $\gamma_t$. In the upper part of each plot, the black dots are the data points with their error bars, and the blue lines are the best-fit predictions from the EFTofLSS presented in this work. The lower part of each plot shows the residuals of the best-fit curves relative to the data diagonal errors (with $y$-axis corresponding to $\pm 3\sigma$). The shaded regions are excluded by the scale cuts used in this analysis; for reference, the DES collaboration scale cut choice DES:2021wwk is shown in dotted vertical lines. For the $\gamma_t$ correlations, the rows (first indices) correspond to the lenses, while the columns (second indices) corresponds to the sources.
• Figure 4: DES Y3 two-point angular correlation functions: cosmic shear $\xi_\pm$. In the upper part of each plot, the black dots are the data points with their error bars, and the blue lines are the best-fit predictions from the EFTofLSS presented in this work. The lower part of each plot shows the residuals of the best-fit curves relative to the data diagonal errors (with $y$-axis corresponding to $\pm 3\sigma$). The shaded regions are excluded by the scale cuts used in this analysis; for reference, the DES collaboration scale cut choice DES:2021wwk is shown in dotted vertical lines.
• Figure 5: Summary of the main contributions entering the predictions of galaxy clustering $w$ and galaxy-galaxy lensing $\gamma_t$ (shown for one source redshift bin only) considered in this work, from the best-fit obtained fitting the Buzzard simulations. The total signal is compared to the individual contributions in the EFTofLSS: linear, one loop, and two loop. The latter, approximated by eq. \ref{['eq:2loop']}, does not enter in our baseline analysis setup and is used to calibrate our choice of scale-cuts. The linear and 1-loop lines correspond to the real-space contribution. The redshift-space distortions and magnification bias, modelled at the linear level only, are explicitly shown. When negative, contributions are shown in dashed lines. The shaded regions are the $1-$ and $2-\sigma$ error bars from DES data, shown for reference.