Field-level Emulation of Cosmic Structure Formation with Cosmology and Redshift Dependence

TL;DR

This paper introduces a neural network-based emulator that accurately models the nonlinear evolution of cosmic structures across different cosmologies and redshifts, improving efficiency and precision in large-scale structure simulations.

Contribution

The authors develop a novel field-level emulator that incorporates cosmology and redshift dependence, predicting nonlinear displacements from linear fields with high accuracy and efficiency.

Findings

Achieves percent-level accuracy at scales of k~1 Mpc/h at z=0

Performs well on unseen cosmologies and redshifts

Provides consistent merger histories compared to N-body simulations

Abstract

We present a field-level emulator for large-scale structure, capturing the cosmology dependence and the time evolution of cosmic structure formation. The emulator maps linear displacement fields to their corresponding nonlinear displacements from N-body simulations at specific redshifts. Designed as a neural network, the emulator incorporates style parameters that encode dependencies on $\Omega_{\rm m}$ and the linear growth factor $D(z)$ at redshift $z$. We train our model on the six-dimensional N-body phase space, predicting particle velocities as the time derivative of the model's displacement outputs. This innovation results in significant improvements in training efficiency and model accuracy. Tested on diverse cosmologies and redshifts not seen during training, the emulator achieves percent-level accuracy on scales of $k\sim~1~{\rm Mpc}^{-1}~h$ at $z=0$, with improved performance at higher redshifts. We compare predicted structure formation histories with N-body simulations via merger trees, finding consistent merger event sequences and statistical properties.