The kinetic Sunyaev Zeldovich effect as a benchmark for AGN feedback models in hydrodynamical simulations: insights from DESI + ACT

TL;DR

This paper demonstrates that baryonic feedback critically reshapes gas distributions and the matter power spectrum, and it leverages the kSZ signal as a direct, high-fidelity benchmark to test and discriminate between hydrodynamical simulation prescriptions. By constructing like-for-like samples using galaxy-galaxy lensing–based halo-mass matching and carefully mitigating cosmic variance and miscentering, the authors show that simulations with strong AGN feedback better reproduce DESI Year 1 + ACT kSZ measurements than fiducial models. The kSZ amplitude correlates with matter-power suppression at $k \,\sim\,1~h\mathrm{Mpc}^{-1}$ and remains informative down to halo masses near $M_{500} \sim 10^{12} M_\odot$, indicating its potential to constrain feedback physics beyond X-ray gas fractions. These results establish the kSZ as a critical tool for validating and tuning baryonic physics in cosmological simulations, with significant implications for small-scale cosmology and the interpretation of Stage-IV weak-lensing data.

Abstract

Baryonic feedback remains one of the largest uncertainties in cosmological hydrodynamical simulations, with different prescriptions producing divergent predictions for the fraction of gas expelled from halos, the radial extent of the gas expulsion and the impact on large scale matter clustering. We present the first systematic study of the kinetic Sunyaev-Zel'dovich (kSZ) effect across a wide range of simulations (FLAMINGO, ANTILLES, BAHAMAS, SIMBA, FABLE and their variants), and compare them directly to DESI Year 1 + ACT kSZ measurements. We ensure a like-for-like comparison with observations by developing a robust methodology that accounts for the halo mass selection using galaxy-galaxy lensing, cosmic variance, miscentering and satellites, establishing the kSZ effect as a new benchmark for the simulations. We find that fiducial feedback models are disfavoured by >3 sigma, while simulations with more powerful AGN feedback within the FLAMINGO and BAHAMAS suites, as well as SIMBA, reproduce the observed kSZ signal within <2 sigma. We use the ANTILLES simulation suite to demonstrate that the amplitude of the kSZ effect is a strong predictor of matter power spectrum suppression, competitive with baryon fraction metrics. These results establish the kSZ as a critical probe for evaluating feedback physics and for advancing the fidelity of cosmological simulations.

Figures (14)

• Figure 1: Left: The matter power spectrum suppression due to baryonic effects, $SP(k)=P(k)/P_{\mathrm{DM only}}(k)$, at $z=0$ (top) and $z=0.75$ (bottom), measured in each of the hydrodynamical simulations we study. Note that we do not measure $SP(k)$ at $z=0.75$ for SIMBA since the corresponding dark matter only simulation output is not available. Centre left: The median total baryon fraction normalised by the universal baryon fraction, $f_{\rm b}/(\Omega_{\rm b}/\Omega_{\rm m})$, as a function of halo mass, $M_{500}$, measured in each of the simulations we study at $z=0$ (top) and $z=0.75$ (bottom). Centre right: The median hot gas mass fraction, $M_{\mathrm{gas}}/M_{500}$, as a function of halo mass, $M_{500}$, measured in each of the simulations we study at $z=0$ (top) and $z=0.75$ (bottom). We measure $M_{\mathrm{gas}}$ within a spherical aperture of radius $R_{500}$ centered on the particle with the minimum gravitational potential energy in the halo. Right: The radial hot gas density profile, showing the gas density enclosed within a radius $R$, plotted as a function of the radius normalized to $R_{500}$. We plot the mean profile computed for halos with $13.25<\log_{10}(M_{500} [\mathrm{M_{\odot}}])<13.35$. In all panels we plot the 400 ANTILLES boxes, as well as FABLE (red), XFABLE (purple), FLAMINGO (navy solid), FLAMINGO$f_{\mathrm{gas}}-8\sigma$ (navy dashed), FLAMINGO jets $f_{\mathrm{gas}}-4\sigma$ (navy dashed), BAHAMAS (blue solid), BAHAMAS 8.0 (blue dashed) and SIMBA (orange). The ANTILLES simulations are colour-coded by the suppression of the matter power spectrum due to baryonic effects measured at $k=1~\mathrm{Mpc}/h$. The pink line represents the ANTILLES box that provides the best combined fit to the DESI + ACT BGS, LRG M1 and LRG M2 kSZ measurements (see Section \ref{['sec:antillescompare']}).
• Figure 2: The impact of cosmic variance on the stacked kSZ profile, measured in sub-volumes of the full $400~\mathrm{Mpc}/h$ BAHAMAS box. In all panels, the colour of each sub-volume profile indicates the RMS line-of-sight velocity of halos in the stack, ${\langle v_{\mathrm{LOS}}^2 \rangle}^{1/2}$. The black dashed line shows the result from the full $400~\mathrm{Mpc}/h$ volume. Upper: halos with mass $13.5<\log_{10}(M_{500}/\mathrm{M_{\odot}})<14$ measured in 64 $100~\mathrm{Mpc}/h$ sub-volumes. Centre: halos with mass $13.5<\log_{10}(M_{500}/\mathrm{M_{\odot}})<14$ measured in eight $200~\mathrm{Mpc}/h$ sub-volumes. Lower: halos with mass $13<\log_{10}(M_{500}/\mathrm{M_{\odot}})<13.5$ measured in 64 $100~\mathrm{Mpc}/h$ sub-volumes.
• Figure 3: The impact of a centrals-only galaxy selection (dashdotted lines), versus a selection including all galaxies and therefore a fraction of satellite galaxies (solid lines), for BGS-like samples at $z \approx 0.3$ with fixed mean halo mass of $\log_{10}\left(\langle M_{500}[\mathrm{M_{\odot}}] \rangle\right) = 13.30$. We show FLAMINGO (navy), SIMBA (orange), BAHAMAS (blue) and FABLE (red).
• Figure 4: The halo mass dependence of the stacked kSZ radial profile in hydrodynamical simulations at $z=0.75$. We measure the kSZ signal within a CAP filter centered on groups and clusters, stacking the profiles in four $M_{500}$ bins. We plot the 400 ANTILLES boxes, as well as FABLE (red), XFABLE (purple), FLAMINGO (navy solid), FLAMINGO$f_{\mathrm{gas}}-8\sigma$ (navy dashed), FLAMINGO jets $f_{\mathrm{gas}}-4\sigma$ (navy dotted), BAHAMAS (blue solid), BAHAMAS 8.0 (blue dashed) and SIMBA (orange). The ANTILLES simulations are colour-coded by the suppression of the matter power spectrum due to baryonic effects measured at $z=0,\ k=1~\mathrm{Mpc}/h$. To better distinguish between simulations, the residual panel shows the difference in the kSZ profile of each simulation relative to FLAMINGO.
• Figure 5: Upper panels: the stacked radial kSZ profile in hydrodynamical simulations in comparison to the DESI Y1 + ACT $z\approx0.3$ BGS and $z\approx0.75$ full sample LRG measurements Ried2025. We shade the measurements at large radii in grey, since they are correlated and therefore effectively do not provide independent constraints. We measure the kSZ signal in simulations within a CAP filter centered on galaxies, with galaxy samples constructed using a stellar mass–based selection, ensuring that the mean halo masses, $\log_{10}(\langle M_{\rm 500}[\mathrm{M_{\odot}}]\rangle)$, match those estimated from the galaxy–galaxy lensing measurements reported in siegel2025b. We plot FABLE (red), XFABLE (purple), FLAMINGO (navy solid), FLAMINGO$f_{\mathrm{gas}}-8\sigma$ (navy dashed), FLAMINGO jets $f_{\mathrm{gas}}-4\sigma$ (navy dotted), BAHAMAS (blue solid), BAHAMAS 8.0 (blue dashed) and SIMBA (orange). We show a shaded region for the 100 Mpc/$h$ boxes, representing the span in the measured kSZ profile attained by projecting along each of the three independent box axes, to demonstrate the impact of cosmic variance. Lower panels: the difference between the simulation predicted kSZ signal and the DESI Y1 + ACT measurements, normalised by the reported DESI Y1 + ACT error bars.