Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userEfficient exploration and calibration of a semi-analytical model of galaxy formation with deep learning
65
0
0.0
(
0
)
Ask ChatGPT about the research
No Arabic abstract
We implement a sample-efficient method for rapid and accurate emulation of semi-analytical galaxy formation models over a wide range of model outputs. We use ensembled deep learning algorithms to produce a fast emulator of an updated version of the GALFORM model from a small number of training examples. We use the emulator to explore the models parameter space, and apply sensitivity analysis techniques to better understand the relative importance of the model parameters. We uncover key tensions between observational datasets by applying a heuristic weighting scheme in a Markov chain Monte Carlo framework and exploring the effects of requiring improved fits to certain datasets relative to others. Furthermore, we demonstrate that this method can be used to successfully calibrate the model parameters to a comprehensive list of observational constraints. In doing so, we re-discover previous GALFORM fits in an automatic and transparent way, and discover an improved fit by applying a heavier weighting to the fit to the metallicities of early-type galaxies. The deep learning emulator requires a fraction of the model evaluations needed in similar emulation approaches, achieving an out-of-sample mean absolute error at the knee of the K-band luminosity function of 0.06 dex with less than 1000 model evaluations. We demonstrate that this is an extremely efficient, inexpensive and transparent way to explore multi-dimensional parameter spaces, and can be applied more widely beyond semi-analytical galaxy formation models.
rate research
Read More
We compare the mean mass assembly histories of compact and fossil galaxy groups in the Millennium dark matter simulation and an associated semi-analytic galaxy formation model. Tracing the halo mass of compact groups (CGs) from z=0 to z=1 shows that, on average, 55 per cent of the halo mass in compact groups is assembled since z~1, compared to 40 per cent of the halo mass in fossil groups (FGs) in the same time interval, indicating that compared to FGs, CGs are relatively younger galaxy systems. At z=0, for a given halo mass, fossil groups tend to have a larger concentration than compact groups. Investigating the evolution of CGs parameters show that they become more compact with time. CGs at z=0.5 see their magnitude gaps increase exponentially, but it takes ~10 Gyr for them to reach a magnitude gap of 2 magnitudes. The slow growth of the magnitude gap leads to only a minority (~41 per cent) of CGs selected at z=0.5 turning into a FG by z=0. Also, while three-quarters of FGs go through a compact phase, most fail to meet the CG isolation criterion, leaving only ~30 per cent of FGs fully satisfying the CG selection criteria. Therefore, there is no strong link of CGs turning into FGs or FGs originating from CGs. The relation between CGs and FGs is thus more complex, and in most cases, FGs and CGs follow different evolutionary tracks.
In a Universe where AGN feedback regulates star formation in massive galaxies, a strong correlation between these two quantities is expected. If the gas causing star formation is also responsible for feeding the central black hole, then a positive correlation is expected. If powerful AGNs are responsible for the star formation quenching, then a negative correlation is expected. Observations so far have mainly found a mild correlation or no correlation at all (i.e. a flat relation between star formation rate (SFR) and AGN luminosity), raising questions about the whole paradigm of AGN feedback. In this paper, we report the predictions of the GALFORM semi-analytical model, which has a very strong coupling between AGN activity and quenching of star formation. The predicted SFR-AGN luminosity correlation appears negative in the low AGN luminosity regime, where AGN feedback acts, but becomes strongly positive in the regime of the brightest AGN. Our predictions reproduce reasonably well recent observations by Rosario et al., yet there is some discrepancy in the normalisation of the correlation at low luminosities and high redshifts. Though this regime could be strongly influenced by observational biases, we argue that the disagreement could be ascribed to the fact that GALFORM neglects AGN variability effects. Interestingly, the galaxies that dominate the regime where the observations imply a weak correlation are massive early-type galaxies that are subject to AGN feedback. Nevertheless, these galaxies retain high enough molecular hydrogen contents to maintain relatively high star formation rates and strong infrared emission.
We present a simple semi-numerical model designed to explore black hole growth and galaxy evolution. This method builds on a previous model for black hole accretion that uses a semi-numerical galaxy formation model and universal Eddington ratio distribution to describe the full AGN population by independently connecting galaxy and AGN growth to the evolution of the host dark matter halos. We fit observed X-ray luminosity functions up to a redshift of z ~ 4, as well as investigate the evolution of the Eddington ratio distributions. We find that the Eddington ratio distribution evolves with redshift such that the slope of the low-Eddington accretion rate distribution increases with cosmic time, consistent with the behavior predicted in hydrodynamical simulations for galaxies with different gas fractions. We also find that the evolution of our average Eddington ratio is correlated with observed star formation histories, supporting a picture in which black holes and galaxies evolve together in a global sense. We further confirm the impact of luminosity limits on observed galaxy and halo properties by applying selection criteria to our fiducial model and comparing to surveys across a wide range of redshifts.
Compact groups (CGs) of galaxies are defined as isolated and dense galaxy systems that appear to be a unique site of multiple galaxy interactions. Semi-analytical models of galaxy formation (SAMs) are a prime tool to understand CGs. We investigate how the frequency and the three-dimensional nature of CGs depends on the SAM and its underlying cosmological parameters. Extracting 9 lightcones of galaxies from 5 different SAMs and selecting CGs as in observed samples, we find that the frequency and nature of CGs depends strongly on the cosmological parameters. Moving from the WMAP1 to the WMAP7 and Planck cosmologies (increasing density of the Universe and decreasing normalisation of the power spectrum), the space density of CGs is decreased by a factor 2.5, while the fraction of CGs that are physically dense falls from 50 to 35 percent. The lower $sigma_8$ leads to fewer dense groups, while the higher $Omega_{rm m}$ causes more chance alignments. However, with increased mass and spatial resolution, the fraction of CGs that are physically dense is pushed back up to 50 percent. The intrinsic differences in the SAM recipes also lead to differences in the frequency and nature of CGs, particularly those related to how SAMs treat orphan galaxies. We find no dependence of CG properties on the flux limit of the mock catalogues nor on the waveband in which galaxies are selected. One should thus be cautious when interpreting a particular SAM for the frequency and nature of CGs.
A semi-analytic model is proposed that couples the Press-Schechter formalism for the number of galaxies with a prescription for galaxy-galaxy interactions that enables to follow the evolution of galaxy morphologies along the Hubble sequence. Within this framework, we calculate the chemo-spectrophotometric evolution of galaxies to obtain spectral energy distributions. We find that such an approach is very successful in reproducing the statistical properties of galaxies as well as their time evolution. We are able to make predictions as a function of galaxy type: for clarity, we restrict ourselves to two categories of galaxies: early and late types that are identified with ellipticals and disks. In our model, irregulars are simply an early stage of galaxy formation. In particular, we obtain good matches for the galaxy counts and redshift distributions of sources from UV to submm wavelengths. We also reproduce the observed cosmic star formation history and the diffuse background radiation, and make predictions as to the epoch and wavelength at which the dust-shrouded star formation of spheroids begins to dominate over the star formation that occurs more quiescently in disks. A new prediction of our model is a rise in the FIR luminosity density with increasing redshift, peaking at about $zsim 3$, and with a ratio to the local luminosity density $rho_{L, u} (z = z_{peak})/ rho_{L, u} (z = 0)$ about 10 times higher than that in the blue (B-band) which peaks near $zsim 2$.
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
