Non-linear galaxy clustering in modified gravity cosmologies

Abstract

We present , a code developed for the very fast production of full -body cosmological simulations in modified gravity (MG) models. We describe the implementation, numerical tests, and first results of a large suite of cosmological simulations for a wide range of viable MG models. The code is highly optimised, with a tremendous speedup of a factor of more than a hundred compared with earlier -body codes, while still giving accurate predictions of the matter power spectrum and dark matter halo abundance. is ideal for the generation of large numbers of MG simulations that can be used in the construction of mock galaxy catalogues and the production of accurate emulators for ongoing and future galaxy surveys.



The coming generation of galaxy surveys will provide measurements of galaxy clustering with unprecedented accuracy and data size, which will allow us to test cosmological models at a much higher precision than previously achievable.
This means that we must have more accurate theoretical predictions to compare with future observational data.
As a first step toward more accurate modelling of the redshift space distortions (RSD) of small-scale galaxy clustering in modified gravity cosmologies, we investigate the validity of the so-called Skew-T (ST) probability distribution function (PDF) of halo pairwise peculiar velocities in these models. We show that combined with the streaming model, the ST PDF substantially improves the small-scale predictions by incorporating skewness and kurtosis, for both cold dark matter (CDM) and two leading MG models: gravity and the DGP braneworld model. The ST model reproduces the velocity PDF and redshift-space halo clustering measured from MG -body simulations down to highly non-linear scales. By performing a simple Fisher analysis, we find a significant increase in constraining power to detect modifications of General Relativity by introducing small-scale information in the RSD analyses.


We introduce an emulator-based halo model approach for non-linear clustering of galaxies in modified gravity cosmologies. We construct accurate emulators, i.e. simulation-based theoretical templates, using neural networks for basic halo properties than can be calculated robustly from -body simulations. The dark matter halo emulators can be combined with a halo-galaxy connection model to predict the galaxy clustering statistics down to non-linear scales through the halo model.