We present the Cosmology and Astrophysics with MachinE Learning Simulations (CAMELS) project. CAMELS is a suite of 4233 cosmological simulations of 25 h − 1 Mpc 3 volume each: 2184 state-of-the-art (magneto)hydrodynamic simulations run with the AREPO and GIZMO codes, employing the same baryonic subgrid physics as the IllustrisTNG and SIMBA simulations, and 2049 N-body simulations. The goal of the CAMELS project is to provide theory predictions for different observables as a function of cosmology and astrophysics, and it is the largest suite of cosmological (magneto)hydrodynamic simulations designed to train machine-learning algorithms. CAMELS contains thousands of different cosmological and astrophysical models by way of varying Ω m , σ 8, and four parameters controlling stellar and active galactic nucleus feedback, following the evolution of more than 100 billion particles and fluid elements over a combined volume of ( 400 h − 1 Mpc ) 3 . We describe the simulations in detail and characterize the large range of conditions represented in terms of the matter power spectrum, cosmic star formation rate density, galaxy stellar mass function, halo baryon fractions, and several galaxy scaling relations. We show that the IllustrisTNG and SIMBA suites produce roughly similar distributions of galaxy properties over the full parameter space but significantly different halo baryon fractions and baryonic effects on the matter power spectrum. This emphasizes the need for marginalizing over baryonic effects to extract the maximum amount of information from cosmological surveys. We illustrate the unique potential of CAMELS using several machine-learning applications, including nonlinear interpolation, parameter estimation, symbolic regression, data generation with Generative Adversarial Networks, dimensionality reduction, and anomaly detection.
We explore the impact of incorporating physically motivated ionisation and recombination rates on the history and topology of cosmic reionisation and the resulting 21-cm power spectrum, by incorporating inputs from small-volume hydrodynamic simulations into our semi-numerical code, SimFast21, that evolves reionisation on large scales. We employ radiative hydrodynamic simulations to parameterize the ionisation rate R ion and recombination rate R rec as functions of halo mass, overdensity and redshift. We find that R ion scales super-linearly with halo mass (R ion ∝ M h), in contrast to previous assumptions. Implementing these scalings into SimFast21, we tune our one free parameter, the escape fraction f esc , to simultaneously reproduce recent observations of the Thomson optical depth, ionizing emissivity, and volumeaveraged neutral fraction by the end of reionisation. This yields f esc = 4 +7 −2 % averaged over our 0.375h −1 Mpc cells, independent of halo mass or redshift, increasing to 6% if we also constrain to match the observed z = 7 star formation rate function. Introducing super-linear R ion increases the duration of reionisation and boosts small-scale 21-cm power by ×2 − 3 at intermediate phases of reionisation, while inhomogeneous recombinations reduce ionised bubble sizes and suppress large-scale 21-cm power by ×2 − 3. Gas clumping on sub-cell scales has a minimal effect on the 21cm power. Super-linear R ion also significantly increases the median halo mass scale for ionising photon output to ∼ 10 10 M , making the majority of reionising sources more accessible to next-generation facilities. These results highlight the importance of accurately treating ionising sources and recombinations for modeling reionisation and its 21-cm power spectrum.
Recent results have suggested that active galactic nuclei (AGN) could provide enough photons to reionise the Universe. We assess the viability of this scenario using a seminumerical framework for modeling reionisation, to which we add a quasar contribution by constructing a Quasar Halo Occupation Distribution (QHOD) based on Giallongo et al. observations. Assuming a constant QHOD, we find that an AGN-only model cannot simultaneously match observations of the optical depth τ e , neutral fraction, and ionising emissivity. Such a model predicts τ e too low by ∼ 2σ relative to Planck constraints, and reionises the Universe at z < ∼ 5. Arbitrarily increasing the AGN emissivity to match these results yields a strong mismatch with the observed ionising emissivity at z ∼ 5. If we instead assume a redshift-independent AGN luminosity function yielding an emissivity evolution like that assumed in Madau & Haardt model, then we can match τ e albeit with late reionisation; however such evolution is inconsistent with observations at z ∼ 4 − 6 and poorly motivated physically. These results arise because AGN are more biased towards massive halos than typical reionising galaxies, resulting in stronger clustering and later formation times. AGN-dominated models produce larger ionising bubbles that are reflected in ∼ ×2 more 21cm power on all scales. A model with equal parts galaxies and AGN contribution is still (barely) consistent with observations, but could be distinguished using next-generation 21cm experiments HERA and SKA-low. We conclude that, even with recent claims of more faint AGN than previously thought, AGN are highly unlikely to dominate the ionising photon budget for reionisation.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.