Confronting cosmic shear astrophysical uncertainties: DES Year 3 revisited

TL;DR

Confronts the limitation of weak-lensing cosmology from baryonic feedback and intrinsic alignments. The authors develop a data-driven framework that combines BCEmu-calibrated baryonification constrained by X-ray gas fractions and kSZ measurements with a Tailored IA model informed by blue galaxies to recover non-linear information from cosmic shear. Applying to DES Year 3 data, they obtain $S_8=0.832^{+0.013}_{-0.017}$ with ~2% precision, a factor of two improvement over fiducial DES Y3, and a $1.5\sigma$ shift toward the CMB value within $\Lambda$CDM. The fit constrains the suppression of the non-linear matter power spectrum at $k=1\,h$/Mpc to $4.3\%$, and indicates stronger feedback than several hydrodynamical simulations. This strategy offers a blueprint for upcoming surveys to unlock the full statistical power of weak lensing by leveraging external gas probes and IA measurements.

Abstract

Cosmology from weak gravitational lensing has been limited by astrophysical uncertainties in baryonic feedback and intrinsic alignments. By calibrating these effects using external data, we recover non-linear information, achieving a 2% constraint on the clustering amplitude, $S_8$, resulting in a factor of two improvement on the $Λ$CDM constraints relative to the fiducial Dark Energy Survey Year 3 model. The posterior, $S_8=0.832^{+0.013}_{-0.017}$, shifts by $1.5σ$ to higher values, in closer agreement with the cosmic microwave background result for the standard six-parameter $Λ$CDM cosmology. Our approach uses a star-forming 'blue' galaxy sample with intrinsic alignment model parameters calibrated by direct spectroscopic measurements, together with a baryonic feedback model informed by observations of X-ray gas fractions and kinematic Sunyaev-Zel'dovich effect profiles that span a wide range in halo mass and redshift. Our results provide a blueprint for next-generation surveys: leveraging galaxy properties to control intrinsic alignments and external gas probes to calibrate feedback, unlocking a substantial improvement in the precision of weak lensing surveys.

Figures (9)

• Figure 1: Marginalised $\Omega_{\rm m}-S_8$ posteriors (left) and the suppression of the matter power spectrum due to baryonic feedback (right) constrained using DES Y3 reanalysed with calibrated feedback and tailored IA (teal). Left: Our result shows a factor of $\times$2 improvement in $S_8$ uncertainty compared to the fiducial DES Y3 model (black; all galaxies, using scale cuts to mitigate baryonic feedback and the TATT IA model) and is in better agreement with the CMB $\Lambda$CDM constraint measured by ACT+Planck TTTEEE+lensing likelihood ACTDR6ACT-ext. Right: By retaining the full scale extent of the cosmic shear measurements, we constrain the matter power spectrum suppression, which, dominated by our informed prior, favors stronger feedback compared to hydrodynamical simulations; FLAMINGO & FLAMINGO $f_{\mathrm{gas}}-8\sigma$Schaye2023, SIMBA Dave2019, BAHAMAS McCarthy2017, MTNG740 Pakmor2023 and (X)FABLE Henden2018XFABLE.
• Figure 2: Marginalized $\Omega_{\rm m}-S_8$ posteriors constrained using DES Y3, showing the impact of each model improvement compared to the fiducial DES choices (TATT IA model with scale cuts, black): Tailored-IA with scale cuts (navy dashed); with an uninformed feedback model without scale cuts (blue, dot-dashed); and with calibrated feedback (teal).
• Figure 3: Marginalised posteriors on $S_8$ and the IA amplitude in four tomographic bin, modelled using NLA. We show constraints from the DES Y3 blue galaxy sample using tailored-IA and calibrated feedback, with the tailored-IA priors shown in grey (Section \ref{['sec:tailorIA']}, McCullough, Siegel et al., in prep.).
• Figure 4: Left: Marginalised posteriors on $S_8$ and BCEmu parameters constrained using the DES Y3 blue galaxy sample with uncalibrated (blue dash-dot) and calibrated (teal) baryonic feedback, with the calibrated feedback prior of Siegel2025_BC shown in grey (Section \ref{['sec:BFmodel']}, grey). Both variants are analyzed with tailored-IA. Right: Marginalised posterior on the suppression of the matter power spectrum due to baryon feedback with the calibrated prior (teal). The baryon feedback prior of Siegel2025_BC (fixed cosmology) is shown in purple and the combination of the feedback prior and wide cosmology prior is shown in grey.
• Figure 5: The effective prior on $\Omega_{\rm m}$ in the 'Tailored-IA; Calibrated feedback' analysis, due to the bcemu emulator failing outside $0.1 < f_b < 0.25$ (grey). The inferred $\Omega_{\rm m}$ posterior attained when evaluating on the data lies within the effective $\Omega_{\rm m}$ prior (teal).