Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userGalaxyNet: Connecting galaxies and dark matter haloes with deep neural networks and reinforcement learning in large volumes
62
0
0.0
(
0
)
Ask ChatGPT about the research
No Arabic abstract
We present the novel wide & deep neural network GalaxyNet, which connects the properties of galaxies and dark matter haloes, and is directly trained on observed galaxy statistics using reinforcement learning. The most important halo properties to predict stellar mass and star formation rate (SFR) are halo mass, growth rate, and scale factor at the time the mass peaks, which results from a feature importance analysis with random forests. We train different models with supervised learning to find the optimal network architecture. GalaxyNet is then trained with a reinforcement learning approach: for a fixed set of weights and biases, we compute the galaxy properties for all haloes and then derive mock statistics (stellar mass functions, cosmic and specific SFRs, quenched fractions, and clustering). Comparing these statistics to observations we get the model loss, which is minimised with particle swarm optimisation. GalaxyNet reproduces the observed data very accurately ($chi_mathrm{red}=1.05$), and predicts a stellar-to-halo mass relation with a lower normalisation and shallower low-mass slope at high redshift than empirical models. We find that at low mass, the galaxies with the highest SFRs are satellites, although most satellites are quenched. The normalisation of the instantaneous conversion efficiency increases with redshift, but stays constant above $zgtrsim0.7$. Finally, we use GalaxyNet to populate a cosmic volume of $(5.9~mathrm{Gpc})^3$ with galaxies and predict the BAO signal, the bias, and the clustering of active and passive galaxies up to $z=4$, which can be tested with next-generation surveys, such as LSST and Euclid.
rate research
Read More
Simple but flexible dynamical models are useful for many purposes, including serving as the starting point for more complex models or numerical simulations of galaxies, clusters, or dark matter haloes. We present SpheCow, a new light-weight and flexible code that allows one to easily explore the structure and dynamics of any spherical model. Assuming an isotropic or Osipkov-Merritt anisotropic orbital structure, the code can automatically calculate the dynamical properties of any model with either an analytical density profile or an analytical surface density profile as starting point. We have extensively validated SpheCow using a combination of comparisons to analytical and high-precision numerical calculations, as well as the calculation of inverse formulae. SpheCow contains readily usable implementations for many standard models, including the Plummer, Hernquist, NFW, Einasto, Sersic, and Nuker models. The code is publicly available as a set of C++ routines and as a Python module, and it is designed to be easily extendable, in the sense that new models can be added in a straightforward way. We demonstrate this by adding two new families of models in which either the density slope or the surface density slope is described by an algebraic sigmoid function. We advocate the use of the SpheCow code to investigate the full dynamical structure for models for which the distribution function cannot be expressed analytically and to explore a much wider range of models than is possible using analytical models alone.
The free electron model with Boltzmann statistics for spherical low-density plasmas (Scientific Reports 9. 20384, 2019) is developed further by numerical calculations with asymptotic relations obtaining the density of electrons, mass densities and the potentials of such plasmas. Solutions are developed as function of a pure number x proportional to the distance from the stellar plasma center (galaxy center) with extremely small coefficient so that these solutions are essentially functions of large astronomical distances and masses. The present plasma is divided into a central part and very long tail where most of the large mass of this plasma is included in the long stellar plasma tail. The present model is specialized to completely ionized Hydrogen plasma where emission and absorption of spectral lines can be neglected in the low density stellar plasma. It is shown that the present low-density plasma might represent dark halo which permeates and surrounds the compact galactic stars. Such plasma is found to be transparent in most of the EM spectrum.
Cosmological simulations play an important role in the interpretation of astronomical data, in particular in comparing observed data to our theoretical expectations. However, to compare data with these simulations, the simulations in principle need to include gravity, magneto-hydrodyanmics, radiative transfer, etc. These ideal large-volume simulations (gravo-magneto-hydrodynamical) are incredibly computationally expensive which can cost tens of millions of CPU hours to run. In this paper, we propose a deep learning approach to map from the dark-matter-only simulation (computationally cheaper) to the galaxy distribution (from the much costlier cosmological simulation). The main challenge of this task is the high sparsity in the target galaxy distribution: space is mainly empty. We propose a cascade architecture composed of a classification filter followed by a regression procedure. We show that our result outperforms a state-of-the-art model used in the astronomical community, and provides a good trade-off between computational cost and prediction accuracy.
Using estimates of dark halo masses from satellite kinematics, weak gravitational lensing, and halo abundance matching, combined with the Tully-Fisher and Faber-Jackson relations, we derive the mean relation between the optical, V_opt, and virial, V_200, circular velocities of early- and late-type galaxies at redshift z~0. For late-type galaxies V_opt ~ V_200 over the velocity range V_opt=90-260 km/s, and is consistent with V_opt = V_maxh (the maximum circular velocity of NFW dark matter haloes in the concordance LCDM cosmology). However, for early-type galaxies V_opt e V_200, with the exception of early-type galaxies with V_opt simeq 350 km/s. This is inconsistent with early-type galaxies being, in general, globally isothermal. For low mass (V_opt < 250 km/s) early-types V_opt > V_maxh, indicating that baryons have modified the potential well, while high mass (V_opt > 400 km/s) early-types have V_opt < V_maxh. Folding in measurements of the black hole mass - velocity dispersion relation, our results imply that the supermassive black hole - halo mass relation has a logarithmic slope which varies from ~1.4 at halo masses of ~10^{12} Msun/h to ~0.65 at halo masses of 10^{13.5} Msun/h. The values of V_opt/V_200 we infer for the Milky Way and M31 are lower than the values currently favored by direct observations and dynamical models. This offset is due to the fact that the Milky Way and M31 have higher V_opt and lower V_200 compared to typical late-type galaxies of the same stellar masses. We show that current high resolution cosmological hydrodynamical simulations are unable to form galaxies which simultaneously reproduce both the V_opt/V_200 ratio and the V_opt-M_star (Tully-Fisher/Faber-Jackson) relation.
The development of methods and algorithms to solve the $N$-body problem for classical, collisionless, non-relativistic particles has made it possible to follow the growth and evolution of cosmic dark matter structures over most of the Universes history. In the best studied case $-$ the cold dark matter or CDM model $-$ the dark matter is assumed to consist of elementary particles that had negligible thermal velocities at early times. Progress over the past three decades has led to a nearly complete description of the assembly, structure and spatial distribution of dark matter haloes, and their substructure in this model, over almost the entire mass range of astronomical objects. On scales of galaxies and above, predictions from this standard CDM model have been shown to provide a remarkably good match to a wide variety of astronomical data over a large range of epochs, from the temperature structure of the cosmic background radiation to the large-scale distribution of galaxies. The frontier in this field has shifted to the relatively unexplored subgalactic scales, the domain of the central regions of massive haloes, and that of low-mass haloes and subhaloes, where potentially fundamental questions remain. Answering them may require: (i) the effect of known but uncertain baryonic processes (involving gas and stars), and/or (ii) alternative models with new dark matter physics. Here we present a review of the field, focusing on our current understanding of dark matter structure from $N$-body simulations and on the challenges ahead.
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
