Fully non-linear simulations of galaxy intrinsic alignments for weak lensing with the MillenniumTNG light-cone

ABSTRACT We present a forward model of a realistic weak lensing galaxy catalogue based on the $740\, \rm {Mpc}$ hydrodynamical millenniumtng (MTNG) simulation. Starting from a full particle and cell light-cone covering one octant of the sky ($0< z< 1.5$), we apply a group and subhalo finder to generate a galaxy catalogue for a fiducial observer. For each galaxy, we compute both intrinsic and lensing-induced shear. The intrinsic component is derived from the luminosity-weighted inertia tensor of stellar particles, while the extrinsic shear is obtained through full-sky ray-tracing. This enables direct predictions of intrinsic alignment (IA) effects on the shear correlation function and convergence statistics in a fully non-linear forward model. We find that IA modifies the convergence power spectrum by up to 20 per cent, alters the probability distribution function tails by 10 per cent–20 per cent, and distorts peak and minimum counts up to 30 per cent, depending on redshift and scale. Cross-redshift correlations reveal additional suppressions in power spectra and non-Gaussian statistics at 10 per cent–30 per cent. The IA signal also affects the shear correlation function, increasing with redshift and galaxy stellar mass. With the highest stellar mass cut applied ($5\times 10^{10}\, \mathrm{ M}_\odot$), the intrinsic shear autocorrelation can rival the gravitational shear on small scales. Our results underscore the need for accurate IA modelling in precision weak lensing cosmology with upcoming Stage IV surveys.