Baryonic feedback suppression of the matter power spectrum: a three-parameter fitting formula and its single-parameter reduction

Abstract

Baryonic feedback can suppress matter clustering by $\sim 10\%$ at $k\sim 1h$/Mpc and $z\sim 1$, and is therefore a major source of systematic errors in weak lensing cosmology. We investigate this effect through the ratio $S(k,z)\equiv P_{\rm hydro}(k,z)/P_{\rm DMO}(k,z)$ measured in the FLAMINGO, IllustrisTNG, and Illustris hydrodynamical simulation suites, together with the one-parameter-at-a-time (1P) variation runs of CAMELS-TNG and CAMELS-SIMBA. We present a three-parameter fitting formula that reproduces all but one of these runs with maximum error less than $0.03$ (and typical error less than $0.01$) at $k\leq 3h/$Mpc and $z\leq 2$; the only exception is a run with unrealistically low $Ω_m=0.1$. Each of the three parameters ($B_0$, $n$, and $a$) has a clear physical interpretation: $B_0$ sets the characteristic scale of the suppression at $z=0$, corresponding to a gas particle displacement scale of $B_0^{1/2}$; $n$ sets the suppression floor $(1-Ω_b/Ω_m)^n$; and $a$ sets the redshift $z_B$ at which the suppression peaks. Furthermore, we find that $B_0$ and $z_B$ are significantly correlated for most runs. Also, $n\simeq 1$ for FLAMINGO, TNG and CAMELS-TNG. For these simulations, the description reduces to a single parameter formula with maximum error less than $0.03$. The formulas remain accurate at max$(|ΔS|)<0.03$ (and typical error less than $0.01$) for cosmologies well away from the calibration cosmology, spanning the range $Ω_m\in [0.2,0.5]$, $Ω_b\in [0.029,0.069]$ and $σ_8\in [0.6,1.0]$. These results imply that, despite the diversity of baryonic physics implementations, an accurate description of $S(k,z)$ requires at most three effective degrees of freedom. Therefore the impact of feedback on weak lensing cosmology can in principle be mitigated internally without significant loss of cosmological constraining power.