Euclid preparation. Testing analytic models of galaxy intrinsic alignments in the Euclid Flagship simulation

Abstract

We model intrinsic alignments (IA) in Euclid's Flagship simulation to investigate its impact on Euclid's weak lensing signal. Our IA implementation in the Flagship simulation takes into account photometric properties of galaxies as well as their dark matter host halos. We compare simulations against theory predictions, determining the parameters of two of the most widely used IA models: the Non Linear Alignment (NLA) and the Tidal Alignment and Tidal Torquing (TATT) models. We measure the amplitude of the simulated IA signal as a function of galaxy magnitude and colour in the redshift range $0.1<z<2.1$. We find that both NLA and TATT can accurately describe the IA signal in the simulation down to scales of $6$-$7 \,h^{-1}\,$Mpc. We measure alignment amplitudes for red galaxies comparable to those of the observations, with samples not used in the calibration procedure. For blue galaxies, our constraints are consistent with zero alignments in our first redshift bin $0.1 < z < 0.3$, but we detect a non-negligible signal at higher redshift, which is, however, consistent with the upper limits set by observational constraints. Additionally, several hydrodynamical simulations predict alignment for spiral galaxies, in agreement with our findings. Finally, the evolution of alignment with redshift is realistic and comparable to that determined in the observations. However, we find that the commonly adopted redshift power-law for IA fails to reproduce the simulation alignments above $z=1.1$. A significantly better agreement is obtained when a luminosity dependence is included, capturing the intrinsic luminosity evolution with redshift in magnitude-limited surveys. We conclude that the Flagship IA simulation is a useful tool for translating current IA constraints into predictions for IA contamination of Euclid-like samples.