We examine the properties of the low-redshift circumgalactic medium (CGM) around star-forming and quenched galaxies in the Simba cosmological hydrodynamic simulations, focusing on comparing H i and metal line absorption to observations from the COS-Halos and COS-Dwarfs surveys. Halo baryon fractions are generally $\lesssim 50{{\ \rm per\ cent}}$ of the cosmic fraction due to stellar feedback at low masses, and jet-mode AGN feedback at high masses. Baryons and metals in the CGM of quenched galaxies are $\gtrsim 90{{\ \rm per\ cent}}$ hot gas, while the CGM of star-forming galaxies is more multi-phase. Hot CGM gas has low metallicity, while warm and cool CGM gas have metallicity close to that of galactic gas. Equivalent widths, covering fractions and total path absorption of H i and selected metal lines (Mg ii, Si iii, C iv and O vi) around a matched sample of Simba star-forming galaxies are mostly consistent with COS-Halos and COS-Dwarfs observations to $\lesssim 0.4$ dex, depending on ion and assumed ionising background. Around matched quenched galaxies, absorption in all ions is lower, with H i absorption significantly under-predicted. Metal-line absorption is sensitive to choice of photo-ionising background; assuming recent backgrounds, Simba matches O vi but under-predicts low ions, while an older background matches low ions but under-predicts O vi. Simba reproduces the observed dichotomy of O vi absorption around star forming and quenched galaxies. CGM metals primarily come from stellar feedback, while jet-mode AGN feedback reduces absorption particularly for lower ions.

We analyze clustering measurements of BOSS galaxies using a simulation-based emulator of two-point statistics. We focus on the monopole and quadrupole of the redshift-space correlation function, and the projected correlation function, at scales of 0.1 ∼ 60 h −1 Mpc. Although our simulations are based on wCDM with general relativity (GR), we include a scaling parameter of the halo velocity field, γ f , defined as the amplitude of the halo velocity field relative to the GR prediction. We divide the BOSS data into three redshift bins. After marginalizing over other cosmological parameters, galaxy bias parameters, and the velocity scaling parameter, we find f σ 8(z = 0.25) = 0.413 ± 0.031, f σ 8(z = 0.4) = 0.470 ± 0.026, and f σ 8(z = 0.55) = 0.396 ± 0.022. Compared with Planck observations using a flat Lambda cold dark matter model, our results are lower by 1.9σ, 0.3σ, and 3.4σ, respectively. These results are consistent with other recent simulation-based results at nonlinear scales, including weak lensing measurements of BOSS LOWZ galaxies, two-point clustering of eBOSS LRGs, and an independent clustering analysis of BOSS LOWZ. All these results are generally consistent with a combination of γ f 1 / 2 σ 8 ≈ 0.75 . We note, however, that the BOSS data is well fit assuming GR, i.e., γ f = 1. We cannot rule out an unknown systematic error in the galaxy bias model at nonlinear scales, but near-future data and modeling will enhance our understanding of the galaxy–halo connection, and provide a strong test of new physics beyond the standard model.

ABSTRACT With the advent of future big-data surveys, automated tools for unsupervised discovery are becoming ever more necessary. In this work, we explore the ability of deep generative networks for detecting outliers in astronomical imaging data sets. The main advantage of such generative models is that they are able to learn complex representations directly from the pixel space. Therefore, these methods enable us to look for subtle morphological deviations which are typically missed by more traditional moment-based approaches. We use a generative model to learn a representation of expected data defined by the training set and then look for deviations from the learned representation by looking for the best reconstruction of a given object. In this first proof-of-concept work, we apply our method to two different test cases. We first show that from a set of simulated galaxies, we are able to detect ${\sim}90{{\ \rm per\ cent}}$ of merging galaxies if we train our network only with a sample of isolated ones. We then explore how the presented approach can be used to compare observations and hydrodynamic simulations by identifying observed galaxies not well represented in the models. The code used in this is available at https://github.com/carlamb/astronomical-outliers-WGAN.

We analyze clustering measurements of BOSS galaxies using a simulation-based emulator of two-point statistics. We focus on the monopole and quadrupole of the redshift-space correlation function, and the projected correlation function, at scales of 0.1 ∼ 60 h −1 Mpc. Although our simulations are based on wCDM with general relativity (GR), we include a scaling parameter of the halo velocity field, γ f , defined as the amplitude of the halo velocity field relative to the GR prediction. We divide the BOSS data into three redshift bins. After marginalizing over other cosmological parameters, galaxy bias parameters, and the velocity scaling parameter, we find f σ 8 (z = 0.25) = 0.404 ± 0.03, f σ 8 (z = 0.4) = 0.444 ± 0.025 and f σ 8 (z = 0.55) = 0.385 ± 0.019. Compared with Planck observations using a flat ΛCDM model, our results are lower by 2.29σ, 1.3σ and 4.58σ respectively. These results are consistent with other recent simulation-based results at non-linear scales, including weak lensing measurements of BOSS LOWZ galaxies, two-point clustering of eBOSS LRGs, and an independent clustering analysis of BOSS LOWZ. All these results are generally consistent with a combination of γ 1/2 f σ 8 ≈ 0.75. We note, however, that the BOSS data is well fit assuming GR, i.e. γ f = 1. We cannot rule out an unknown systematic error in the galaxy bias model at non-linear scales, but near-future data and modeling will enhance our understanding of the galaxy-halo connection, and provide a strong test of new physics beyond the standard model.

The problem of anomaly detection in astronomical surveys is becoming increasingly important as data sets grow in size. We present the results of an unsupervised anomaly detection method using a Wasserstein generative adversarial network (WGAN) on nearly one million optical galaxy images in the Hyper Suprime-Cam (HSC) survey. The WGAN learns to generate realistic HSC-like galaxies that follow the distribution of the data set; anomalous images are defined based on a poor reconstruction by the generator and outlying features learned by the discriminator. We find that the discriminator is more attuned to potentially interesting anomalies compared to the generator, and compared to a simpler autoencoder-based anomaly detection approach, so we use the discriminator-selected images to construct a high-anomaly sample of ∼13 000 objects. We propose a new approach to further characterize these anomalous images: we use a convolutional autoencoder to reduce the dimensionality of the residual differences between the real and WGAN-reconstructed images and perform UMAP clustering on these. We report detected anomalies of interest including galaxy mergers, tidal features, and extreme star-forming galaxies. A follow-up spectroscopic analysis of one of these anomalies is detailed in the Appendix; we find that it is an unusual system most likely to be a metal-poor dwarf galaxy with an extremely blue, higher-metallicity H ii region. We have released a catalog with the WGAN anomaly scores; the code and catalog are available at https://github.com/kstoreyf/anomalies-GAN-HSC, and our interactive visualization tool for exploring the clustered data is at https://weirdgalaxi.es.