Search for: All records

Abstract We train graph neural networks to perform field-level likelihood-free inference using galaxy catalogs from state-of-the-art hydrodynamic simulations of the CAMELS project. Our models are rotational, translational, and permutation invariant and do not impose any cut on scale. From galaxy catalogs that only contain 3D positions and radial velocities of ∼1000 galaxies in tiny ( 25 h − 1 Mpc ) 3 volumes our models can infer the value of Ω m with approximately 12% precision. More importantly, by testing the models on galaxy catalogs from thousands of hydrodynamic simulations, each having a different efficiency of supernova and active galactic nucleus feedback, run with five different codes and subgrid models—IllustrisTNG, SIMBA, Astrid, Magneticum, SWIFT-EAGLE—we find that our models are robust to changes in astrophysics, subgrid physics, and subhalo/galaxy finder. Furthermore, we test our models on 1024 simulations that cover a vast region in parameter space—variations in five cosmological and 23 astrophysical parameters—finding that the model extrapolates really well. Our results indicate that the key to building a robust model is the use of both galaxy positions and velocities, suggesting that the network has likely learned an underlying physical relation that does not depend on galaxy formation and is valid on scales larger than ∼10 h −1 kpc. 

In dusty cool-star outflow or ejection events around asymptotic giant branch (AGB) or R Coronae Borealis or RCB-like stars, dust is accelerated by radiation from the star and coupled to the gas via collisional drag forces. It has recently been shown that such dust-gas mixtures are unstable to a super-class of instabilities called the resonant drag instabilities (RDIs), which promote dust clustering. We therefore consider idealized simulations of the RDIs operating on a spectrum of dust grain sizes subject to radiative acceleration (allowing for different grain optical properties), coupled to the gas with a realistic drag law, including or excluding the effects of magnetic fields and charged grains, and calculate for the first time how the RDIs could contribute to observed variability. We show that the RDIs naturally produce significant variations (spatially and temporally) ($\sim 10\!-\!20{{\ \rm per\ cent}}$ 1 σ-level) in the extinction, corresponding to $\sim 0.1\!-\!1\,$mag level in the stellar types above, on time-scales of order months to a year. The fluctuations are surprisingly robust to the assumed size of the source as they are dominated by large-scale modes, which also means their spatial structure could be resolved in some nearby systems. We also quantify how this produces variations in the line-of-sight grain size-distribution. All of these variations are similar to those observed, suggesting that the RDIs may play a key role driving observed spatial and temporal variability in dust extinction within dusty outflow/ejection events around cool stars. We further propose that the measured variations in grain sizes could directly be used to identify the presence of the RDIs in close by systems with observations.

 

Abstract We train graph neural networks on halo catalogs from Gadget N -body simulations to perform field-level likelihood-free inference of cosmological parameters. The catalogs contain ≲5000 halos with masses ≳10 10 h −1 M ⊙ in a periodic volume of ( 25 h − 1 Mpc ) 3 ; every halo in the catalog is characterized by several properties such as position, mass, velocity, concentration, and maximum circular velocity. Our models, built to be permutationally, translationally, and rotationally invariant, do not impose a minimum scale on which to extract information and are able to infer the values of Ω m and σ 8 with a mean relative error of ∼6%, when using positions plus velocities and positions plus masses, respectively. More importantly, we find that our models are very robust: they can infer the value of Ω m and σ 8 when tested using halo catalogs from thousands of N -body simulations run with five different N -body codes: Abacus, CUBEP 3 M, Enzo, PKDGrav3, and Ramses. Surprisingly, the model trained to infer Ω m also works when tested on thousands of state-of-the-art CAMELS hydrodynamic simulations run with four different codes and subgrid physics implementations. Using halo properties such as concentration and maximum circular velocity allow our models to extract more information, at the expense of breaking the robustness of the models. This may happen because the different N -body codes are not converged on the relevant scales corresponding to these parameters. 

Radiation-dust driven outflows, where radiation pressure on dust grains accelerates gas, occur in many astrophysical environments. Almost all previous numerical studies of these systems have assumed that the dust was perfectly coupled to the gas. However, it has recently been shown that the dust in these systems is unstable to a large class of ‘resonant drag instabilities’ (RDIs) which de-couple the dust and gas dynamics and could qualitatively change the non-linear outcome of these outflows. We present the first simulations of radiation-dust driven outflows in stratified, inhomogeneous media, including explicit grain dynamics and a realistic spectrum of grain sizes and charge, magnetic fields and Lorentz forces on grains (which dramatically enhance the RDIs), Coulomb and Epstein drag forces, and explicit radiation transport allowing for different grain absorption and scattering properties. In this paper, we consider conditions resembling giant molecular clouds (GMCs), H ii regions, and distributed starbursts, where optical depths are modest (≲1), single-scattering effects dominate radiation-dust coupling, Lorentz forces dominate over drag on grains, and the fastest-growing RDIs are similar, such as magnetosonic and fast-gyro RDIs. These RDIs generically produce strong size-dependent dust clustering, growing non-linear on time-scales that are much shorter than the characteristic times of the outflow. The instabilities produce filamentary and plume-like or ‘horsehead’ nebular morphologies that are remarkably similar to observed dust structures in GMCs and H ii regions. Additionally, in some cases they strongly alter the magnetic field structure and topology relative to filaments. Despite driving strong micro-scale dust clumping which leaves some gas ‘behind,’ an order-unity fraction of the gas is always efficiently entrained by dust.

 

Abstract The Cosmology and Astrophysics with Machine Learning Simulations (CAMELS) project was developed to combine cosmology with astrophysics through thousands of cosmological hydrodynamic simulations and machine learning. CAMELS contains 4233 cosmological simulations, 2049 N -body simulations, and 2184 state-of-the-art hydrodynamic simulations that sample a vast volume in parameter space. In this paper, we present the CAMELS public data release, describing the characteristics of the CAMELS simulations and a variety of data products generated from them, including halo, subhalo, galaxy, and void catalogs, power spectra, bispectra, Ly α spectra, probability distribution functions, halo radial profiles, and X-rays photon lists. We also release over 1000 catalogs that contain billions of galaxies from CAMELS-SAM: a large collection of N -body simulations that have been combined with the Santa Cruz semianalytic model. We release all the data, comprising more than 350 terabytes and containing 143,922 snapshots, millions of halos, galaxies, and summary statistics. We provide further technical details on how to access, download, read, and process the data at https://camels.readthedocs.io .