Search

As the next generation of large galaxy surveys come online, it is becoming increasingly important to develop and understand the machine-learning tools that analyze big astronomical data. Neural networks are powerful and capable of probing deep patterns in data, but they must be trained carefully on large and representative data sets. We present a new "hump" of the Cosmology and Astrophysics with MachinE Learning Simulations (CAMELS) project: CAMELS-SAM, encompassing one thousand dark-matter-only simulations of (100 h −1 cMpc)3 with different cosmological parameters (Ω m and σ 8) and run through the Santa Cruz semi-analytic model for galaxy formation over a broad range of astrophysical parameters. As a proof of concept for the power of this vast suite of simulated galaxies in a large volume and broad parameter space, we probe the power of simple clustering summary statistics to marginalize over astrophysics and constrain cosmology using neural networks. We use the two-point correlation, count-in-cells, and void probability functions, and we probe nonlinear and linear scales across 0.68 < R <27 h −1 cMpc. We find our neural networks can both marginalize over the uncertainties in astrophysics to constrain cosmology to 3%–8% error across various types of galaxy selections, while simultaneously learning about the SC-SAM astrophysical parameters. This work encompasses vital first steps toward creating algorithms able to marginalize over the uncertainties in our galaxy formation models and measure the underlying cosmology of our Universe. CAMELS-SAM has been publicly released alongside the rest of CAMELS, and it offers great potential to many applications of machine learning in astrophysics: https://camels-sam.readthedocs.io. [ABSTRACT FROM AUTHOR]

We use large simulations of Ly α emitters with different fractions of ionized intergalactic medium to quantify the clustering of Ly α emitters as measured by the void probability function (VPF), and how it evolves under different ionization scenarios. We quantify how well we might be able to distinguish between these scenarios with a deep spectroscopic survey using the future Nancy Grace Roman Space Telescope. Because Roman will be able to carry out blind spectroscopic surveys of Ly α emitters continuously between 7 < z < 12 to sensitivities of at least 10−17 erg s−1 over a wide field of view, it can measure the epoch of reionization as well as the pace of ionization of the intergalactic medium (IGM). We compare deep Roman surveys covering roughly 1, 4, and 16 deg2, and quantify what constraints on reionization the VPF may find for these surveys. A survey of 1 deg2 would distinguish between very late reionization and early reionization to 3 σ near z = 7.7 with the VPF. The VPF of a 4 deg2 survey can distinguish between slow versus fast, as well as early versus late, reionization at >3 σ –4 σ at several redshifts between 7 < z < 9. However, a survey of 13–16 deg2 would allow the VPF to give several robust constraints (>5–8 σ) across the epoch of reionization, and would yield a detailed history of the reionization of the IGM and its effect on Ly α emitter clustering. [ABSTRACT FROM AUTHOR]

We present a new measurement of the Ly α luminosity function (LF) at redshift z = 6.9, finding moderate evolution from z = 5.7 that is consistent with a fully or largely ionized z ∼ 7 intergalactic medium. Our result is based on four fields of the LAGER (Lyman Alpha Galaxies in the Epoch of Reionization) project. Our survey volume of 6.1 × 106 Mpc3 is double that of the next largest z ∼ 7 survey. We combine two new LAGER fields (WIDE12 and GAMA15A) with two previously reported LAGER fields (COSMOS and CDFS). In the new fields, we identify N = 95 new z = 6.9 Ly α emitter (LAEs) candidates, characterize our survey’s completeness and reliability, and compute Ly α LFs. The best-fit Schechter LF parameters for all four LAGER fields are in good general agreement. Two fields (COSMOS and WIDE12) show evidence for a bright-end excess above the Schechter function fit. We find that the Ly α luminosity density declines at the same rate as the UV continuum LF from z = 5.7 to 6.9. This is consistent with an intergalactic medium that was fully ionized as early as redshift z ∼ 7 or with a volume-averaged neutral hydrogen fraction of x H I < 0.33 at 1 σ. [ABSTRACT FROM AUTHOR]

We calculate the void probability function (VPF) in simulations of Lyα emitters (LAEs) across a wide redshift range (z = 3.1, 4.5, 5.7, 6.6). The VPF measures the zero-point correlation function (i.e., places devoid of galaxies) and naturally connects to higher-order correlation functions while being computationally simple to calculate. We explore the Poissonian and systematic errors on the VPF, specify its accuracy as a function of average source density and the volume probed, and provide the appropriate size scales to measure the VPF. At small radii, the accuracy of the VPF is limited by galaxy density, while at large radii, the VPF is limited by the number of independent volumes probed. We also offer guidelines for understanding and quantifying the error in the VPF. We approximate the error in the VPF by using independent subvolumes of the catalogs, after finding that jackknife statistics underestimate the uncertainty. We use the VPF to probe the strength of higher-order correlation functions by measuring and examining the hierarchical scaling between the correlation functions using the count-in-cells method. The negative binomial model (NBM) has been shown to best describe the scaling between the two-point correlation function and the VPF for low-redshift galaxy observations. We further test the fit of the NBM by directly deriving the volume-averaged two-point correlation function from the VPF and vice versa. We find the NBM best describes the simulated LAEs, with a 1σ deviation from the model in the z = 6.6 catalog. This suggests that LAEs show higher-order clustering terms similar to those of normal low-redshift galaxies. [ABSTRACT FROM AUTHOR]