Fast production of cosmological emulators in modified gravity: the matter power spectrum

TL;DR

This paper develops a fast, accurate neural network emulator for the matter power spectrum boost factor in modified gravity theories, trained on COLA simulations and validated against MG-AREPO data, to aid high-redshift galaxy survey analysis.

Contribution

It introduces a neural network emulator for the DGP gravity boost factor, trained on COLA simulations, with a focus on efficiency and accuracy for high-redshift cosmological studies.

Findings

Emulator achieves ~3% accuracy up to k=5 h/Mpc.

Minimal simulation requirements identified for accurate boost factor predictions.

Emulator publicly available for community use.

Abstract

We test the convergence of fast simulations based on the COmoving Lagrangian Acceleration (COLA) method for predictions of the matter power spectrum, specialising our analysis in the redshift range $1 \le z \le 1.65$, relevant to high-redshift spectroscopic galaxy surveys. We then focus on the enhancement of the matter power spectrum in modified gravity (MG), the boost factor, using the Dvali-Gabadadze-Porrati (DGP) theory as a test case but developing a general approach that can be applied to other MG theories. After identifying the minimal simulation requirements for accurate DGP boost factors, we design and produce a COLA simulation suite that we use to train a neural network emulator for the DGP boost factor. Using MG-AREPO simulations as a reference, we estimate the emulator accuracy to be of $\sim 3\%$ up to $k=5 \, h {\rm Mpc}^{-1}$ at $0 \leq z \leq 2$. We make the emulator publicly available at: https://github.com/BartolomeoF/nDGPemu.