Galaxy formation in the brane world I: overview and first results

Abstract

We carry out “full-physics” hydrodynamical simulations of galaxy formation in the normal-branch Dvali-Gabadadze-Porrati (nDGP) braneworld model using a new modified version of the Arepo code and the IllustrisTNG galaxy formation model. We simulate two nDGP models (N5 and N1) which represent, respectively, weak and moderate departures from GR, in boxes of sizes 62 h−1Mpc and 25 h−1Mpc using 2 × 5123 dark matter particles and initial gas cells. This allows us to explore, for the first time, the impact of baryonic physics on galactic scales in braneworld models of modified gravity and to make predictions on the stellar content of dark matter haloes and galaxy evolution through cosmic time in these models. We find significant differences between the GR and nDGP models in the power spectra and correlation functions of gas, stars and dark matter of up to ∼25 per cent on large scales. Similar to their impact in the standard cosmological model (ΛCDM), baryonic effects can have a significant influence over the clustering of the overall matter distribution, with a sign that depends on scale. Studying the degeneracy between modified gravity and galactic feedback in these models, we find that these two physical effects on matter clustering can be cleanly disentangled, allowing for a method to accurately predict the matter power spectrum with baryonic effects included, without having to run hydrodynamical simulations. Depending on the braneworld model, we find differences compared with GR of up to ∼15 per cent in galaxy properties such as the stellar-to-halo-mass ratio, galaxy stellar mass function, gas fraction and star formation rate density. The amplitude of the fifth force is reduced by the presence of baryons in the very inner part of haloes, but this reduction quickly becomes negligible above ∼0.1 times the halo radius.

Fast full $N$-body simulations of generic modified gravity: derivative coupling models

“…The abundance of haloes is enhanced from ≈ 1 to 250 percent for the different nDGP models. The large increase of high-mass haloes in the less efficiently screened nDGP models (models with H 0 r c < 5) is due to the accretion of surrounding matter around these massive objects thanks to the enhanced gravity force: these objects, often being the dominating object within some large surrounding region, can attract matter from the whole region, including the accretion of smaller haloes to them, and so the fifth force can strongly boost their masses; on the other hand, smaller objects, while also experiencing the fifth force [96], are more likely to meet competitors and so their masses grow less.…”

“…As we will demonstrate below, the inclusion of modified gravity solvers in MG-GLAM adds an overhead to the computational cost of GLAM, and for the models considered in this paper and its twin paper [103], a MG-GLAM run can take about 2-5 times (depending on the resolution) the computing time of an equivalent ΛCDM simulation run using default GLAM. All in all, this makes this new code at least around 100 times faster than other modified gravity simulation codes such as ECOSMOG [89][90][91][92] and MG-AREPO [95,96] for the same simulation boxsize and particle number. In spite of such a massive improvement in speed over those latter codes, it is worthwhile to note that MG-GLAM is not an approximate code: it solves the full Poisson and MG equations, and its accuracy is only limited by the resolution of the PM grid used, which can be specified by users based on their particular scientific objectives.…”

“…In particular, to achieve the high level of precision required by galaxy surveys, one can generate hundreds or thousands of independent galaxy mocks that cover the expected survey volume, based on these simulations. However, this has so far been impossible for MG models, which usually involve highly non-linear partial differential equations that govern the new physics, solving which has proven to be very expensive even with the latest codes, e.g., ECOSMOG [89][90][91][92], MG-GADGET [93], ISIS [94] and MG-AREPO [95,96] (see [97] for a comparison of several MG codes). For example, current MG simulations can take between 2 to O (10) times longer than standard ΛCDM simulations of the same specifications.…”

Fast full N-body simulations of generic modified gravity: conformal coupling models

“…As we will demonstrate below, the inclusion of modified gravity solvers in -adds an overhead to the computational cost of , and for the models considered in this paper and its twin paper [102], a -run takes about 3-5 times (depending on the resolution) the computing time of an equivalent ΛCDM simulation using default . All in all, this makes this new code at least 100 times faster than other modified gravity simulation codes such as [88][89][90][91] and - [94,95] for the same simulation boxsize and particle number. In spite of such a massive improvement in speed over those latter codes, it is worthwhile to note that -is not an approximate code: it solves the full Poisson and MG equations, and its accuracy is only limited by the resolution of the PM grid used, which can be specified by users based on their particular scientific objectives.…”

“…In particular, to achieve the high level of precision required by galaxy surveys, one can generate hundreds or thousands of independent galaxy mocks that cover the expected survey volume, using these simulations. However, this has so far been impossible for MG models, which usually involve highly non-linear partial differential equations that govern the new physics, solving which has proven to be very expensive even with the latest codes, e.g., [88][89][90][91], - [92], [93] and - [94,95] (see [96] for a comparison of several MG…”