The Bacco Simulation Project: Bacco Hybrid Lagrangian Bias Expansion Model in Redshift Space

Abstract

We present an emulator that accurately predicts the power spectrum of galaxies in redshift space as a function of cosmological parameters. Our emulator is based on a 2nd-order Lagrangian bias expansion that is displaced to Eulerian space using cosmological $N$-body simulations. Redshift space distortions are then imprinted using the non-linear velocity field of simulated particles and haloes. We build the emulator using a forward neural network trained with the simulations of the BACCO project, which covers an 8-dimensional parameter space including massive neutrinos and dynamical dark energy. We show that our emulator provides unbiased cosmological constraints from the monopole, quadrupole, and hexadecapole of a mock galaxy catalogue that mimics the BOSS-CMASS sample down to nonlinear scales ($k\sim0.6$[$h/$Mpc]$^{3}$). This work opens up the possibility of robustly extracting cosmological information from small scales using observations of the large-scale structure of the Universe.