Baryonic Imprints on DM Halos: characterizing the full concentration-mass probability distribution with CAMELS
The halo concentration, which represents the distribution of matter within a halo, is a vital tool for modeling the matter distribution on scales ranging from megaparsecs to sub-kiloparsecs. While significant progress has been made on connecting the impact of astrophysics on the mean concentration as a function of halo mass, the same connection to higher moments of the concentration distribution remains far less explored. We employ the IllustrisTNG runs in the CAMELS simulation suite to simultaneously quantify the astrophysics and cosmology dependence of the first four moments of the concentration-mass relation; the mean, the scatter, the skewness, and the kurtosis. We use normalizing flows to model the concentration distribution as a function of halo mass, redshift, and six cosmological/astrophysical parameters. We focus on the mass range $M_{\rm vir} \in [10^{11}, 10^{14.5}] M_{\odot}/h$, and the redshift range $z \in [0, 1]$. All higher-order moments exhibit a mass-dependent sensitivity to both astrophysics and cosmology. The behavior in dark matter-only simulations follows the known effects of active mergers and of tidal stripping in smaller halos. Once galaxy formation is included, the scatter shows non-monotonic behavior at the Milky Way (MW) mass-scale, and the amplitude of this behavior is strongly influenced by the baryon fraction and supernovae (SNe) feedback parameters. The skewness has a large sensitivity to the baryon fraction as well and is mostly insensitive to the other parameters, while the kurtosis follows the behavior set by the skewness. Results for the ASTRID simulation suite show similar behavior as that of IllustrisTNG. The cosmology and astrophysics dependence of these higher-order moments impact analyses that are sensitive to the full distribution of the concentration-mass relation, particularly for halos at/below the MW mass scale.