Cosmological Constraints from Dark Energy Survey Year 1 Cluster Lensing and Abundances with Simulation-based Forward-Modeling

TL;DR

The paper develops a simulation-based forward-modeling framework to extract cosmological information from optical galaxy clusters, explicitly embedding cosmology-dependent selection, projection effects, miscentering, and baryonic feedback into a fully simulated inference pipeline. Using DES-Y1 redMaPPer clusters and AbacusSummit simulations, the authors map counts-in-cylinders to redMaPPer richness via abundance matching, model lensing with a compressed baryonification scheme, and employ an emulator plus Gaussian likelihood to constrain $Ω_m$ and $σ_8$ (finding $Ω_m ≈ 0.254$ and $σ_8 ≈ 0.826$ without strong priors, with modest Planck tension). The framework yields competitive, largely consistent results with other low-redshift probes and improves precision by enabling low-richness, small-scale information to inform cosmology. This forward-modeling approach demonstrates the viability of optical cluster cosmology as a powerful tool for precision cosmology in the upcoming Stage IV era, with clear paths to incorporate more realistic galaxy populations and cluster-finding refinements.

Abstract

We present a simulation-based forward-modeling framework for cosmological inference from optical galaxy-cluster samples, and apply it to the abundance and weak-lensing signals of DES-Y1 redMaPPer clusters. The model embeds cosmology-dependent optical selection using a counts-in-cylinders approach, while also accounting for cluster miscentering and baryonic feedback in lensing. Applied to DES-Y1, and assuming a flat $Λ$CDM cosmology, we obtain $Ω_m=0.254^{+0.026}_{-0.020}$ and $σ_8=0.826^{+0.030}_{-0.034}$, consistent with a broad suite of low-redshift structure measurements, including recent full-shape analyses, the DES/KiDS/HSC 3$\times$2 results, and most cluster-abundance studies. Our results are also consistent with \textit{Planck}, with the difference being significant at $2.58σ$. These results establish simulation-based forward-modeling of cluster abundances as a promising new tool for precision cosmology with Stage~IV survey data.

Figures (10)

• Figure 1: The AbacusSummit cosmology sampling in $\Omega_{\rm m}$ and $\sigma_8$ used to train our emulators along with the projection of the AbacusSummit prior ellipsoid in the $\Omega_{\rm m}$-$\sigma_8$ plane. This ellipsoid is meant to represent a conservative (roughly 6--8$\sigma$) constraint on cosmological parameters from CMB+LSS data circa 2021 Maksimova_Summit_et_al_2021. Our work considers two analyses, one that adopts a flat prior within this ellipsoid, which for brevity we refer to as the "AbacusSummit" prior, and one that does not adopt this prior.
• Figure 2: A comparison of Giri_Schneider_2021 baryonification models to our one-parameter approximation. Red and blue correspond to strong and weak baryonic feedback scenarios. The bands are calculated using the full Giri_Schneider_2021 models, with the width of the band corresponding to the $\sim4\%$ uncertainty of our emulator. The solid lines are the best-fit models calculated using our dimensionally compressed 1-parameter model, and are fit to all mass bins we consider simultaneously. Here we only plot the fits for four representative masses. The rms difference between the full model and our one-parameter fits is $\sim2\%$.
• Figure 3: Lensing profiles in redshift bin $z \in [0.20, 0.35)$ (points with errorbars) compared to simulation predictions for our training data (lines); we emphasize that data is included only for reference and that we are not presenting a fit. Colors correspond to richness bins $\lambda \in [60, \infty)$ (red), $[45, 60)$ (yellow), $[30, 45)$ (green), $[20, 30)$ (blue), $[14, 20)$ (purple), and $[10, 14)$ (brown). The bottom panels show the leave-one-out error for each richness bin compared to the observable covariance (band). The solid and dashed lines correspond to the mean and $1\sigma$ errors, respectively.
• Figure 4: Comparison of DES-Y1 cluster weak lensing profiles (point with errorbars) in redshift bins $z\in[0.20, 0.35)$ (left panel), $z\in[0.30, 0.50)$ (middle panel), and $z\in[0.50,0.65)$ (right panel) with those predicted by our fiducial posterior mean model and 200 random samples from our fiducial MCMC chain. The $\chi^2$ of the best fit model is $\chi^2/\mathrm{dof}=190.51/188$.
• Figure 5: $68\%$ and $95\%$ confidence contours in the $\sigma_8$--$\Omega_{\rm m}$ plane for our analysis with (red dashed) and without (orange filled) the AbacusSummit sampling prior (gray). Tensions with other data sets are quoted with respect to the posterior without the AbacusSummit prior. We compare our results with those from the original DES-Y1 analysis of cluster lensing and abundance DESY1CL_2020_et_al, DES-Y1 3$\times$2 DES_3x2pt_2018, and constraints from Planck CMB measurements Planck_DR18_2020. The result of combining our cluster constraints with those from DES-Y1 $3\times2$ is shown in black.