De-baryonifying halos via optimal transport

Abstract

Baryonic feedback uncertainty is a limiting systematic for next-generation weak gravitational lensing analyses. At the same time, high-resolution weak lensing maps are best analyzed at the field-level. Thus, robustly accounting for the baryonic effects in the projected matter density field is required. Ideally, constraints on feedback strength from astrophysical probes should be folded into the weak lensing field-level likelihood. We propose a macroscopic method based on an empirical correlation between feedback strength and an optimal transport cost. Since feedback is local re-distribution of matter, optimal transport is a promising concept. In this proof-of-concept, we de-baryonify projected mass around individual halos in the IllustrisTNG simulation. We choose the de-baryonified solution as the point of maximum likelihood on the hypersurface defined by fixed optimal transport cost around the observed full-physics halos. The likelihood is approximated through a normalizing flow trained on multiple gravity-only simulations. We find that the set of de-baryonified halos reproduces the correct convergence power spectrum suppression. There is considerable scatter when considering individual halos. We outline how the optimal transport de-baryonification concept can be generalized to full convergence maps.