We present a new statistical method to determine the relationship between the stellar masses of galaxies and the masses of their host dark matter haloes over the entire cosmic history from z ∼ 4 to ...the present. This multi-epoch abundance matching (MEAM) model self-consistently takes into account that satellite galaxies first become satellites at times earlier than they are observed. We employ a redshift-dependent parametrization of the stellar-to-halo-mass relation to populate haloes and subhaloes in the Millennium simulations with galaxies, requiring that the observed stellar mass functions at several redshifts are reproduced simultaneously. We show that physically meaningful growth of massive galaxies is consistent with these data only if observational mass errors are taken into account. Using merger trees extracted from the dark matter simulations in combination with MEAM, we predict the average assembly histories of galaxies, separating into star formation within the galaxies (in situ) and accretion of stars (ex situ). Our main results are the peak star formation efficiency decreases with redshift from 23 per cent at z = 0 to 9 per cent at z =4 while the corresponding halo mass increases from 1011.8 to 1012.5 M. The star formation rate of central galaxies peaks at a redshift which depends on halo mass; for massive haloes this peak is at early cosmic times while for low-mass galaxies the peak has not been reached yet. In haloes similar to that of the Milky Way about half of the central stellar mass is assembled after z = 0.7. In low-mass haloes, the accretion of satellites contributes little to the assembly of their central galaxies, while in massive haloes more than half of the central stellar mass is formed ex situ with significant accretion of satellites at z < 2. We find that our method implies a cosmic star formation history and an evolution of specific star formation rates which are consistent with those inferred directly. We present convenient fitting functions for stellar masses, star formation rates and accretion rates as functions of halo mass and redshift.
ABSTRACT
We present constraints on the emergence and evolution of passive galaxies with the empirical model emerge, which reproduces the evolution of stellar mass functions (SMFs), specific and ...cosmic star formation rates since $z$ ≈ 10, ‘quenched’ galaxy fractions, and correlation functions. At fixed halo mass, present-day passive galaxies are more massive than active galaxies, whereas at fixed stellar mass passive galaxies populate more massive haloes in agreement with observations. This effect naturally results from the shape and scatter of the stellar-to-halo mass relation. The stellar mass assembly of present-day passive galaxies is dominated by ‘in situ’ star formation below ∼3 × 1011 M⊙ and by merging and accretion of ‘ex situ’ formed stars at higher mass. The mass dependence is in tension with current cosmological simulations. Lower mass passive galaxies show extended star formation towards low redshift in agreement with IFU surveys. All passive galaxies have main progenitors on the ‘main sequence of star formation’ with the ‘red sequence’ appearing at $z$ ≈ 2. Above this redshift, over 95 per cent of the progenitors of passive galaxies are active. More than 90 per cent of $z$ ≈ 2 ‘main sequence’ galaxies with m* > 1010 M⊙ evolve into present-day passive galaxies. Above redshift 6, more than 80 per cent of the observed SMFs above 109 M⊙ can be accounted for by progenitors of passive galaxies with m* > 1010 M⊙. This implies that high-redshift observations mainly probe the birth of present-day passive galaxies. emerge is available at github.com/bmoster/emerge.
Exoplanet detection using machine learning Malik, Abhishek; Moster, Benjamin P; Obermeier, Christian
Monthly notices of the Royal Astronomical Society,
07/2022, Letnik:
513, Številka:
4
Journal Article
Recenzirano
ABSTRACT
We introduce a new machine learning based technique to detect exoplanets using the transit method. Machine learning and deep learning techniques have proven to be broadly applicable in ...various scientific research areas. We aim to exploit some of these methods to improve the conventional algorithm based approaches presently used in astrophysics to detect exoplanets. Using the time series analysis library tsfresh to analyse light curves, we extracted 789 features from each curve, which capture the information about the characteristics of a light curve. We then used these features to train a gradient boosting classifier using the machine learning tool lightgbm. This approach was tested on K2 campaign 7 data with injected artificial transit signals, which showed that it is competitive compared to the conventional box least-squares fitting method. We further found that our method produced comparable results to existing state-of-the-art deep learning models, while being much more computationally efficient and without needing folded and secondary views of the light curves. For Kepler data, the method is able to predict a planet with an AUC of 0.948, so that 94.8 per cent of the true planet signals are ranked higher than non-planet signals. The resulting recall is 0.96, so that 96 per cent of real planets are classified as planets. For the Transiting Exoplanet Survey Satellite data, we found our method can classify light curves with an accuracy of 0.98, and is able to identify planets with a recall of 0.82 at a precision of 0.63.
ABSTRACT
The James Webb Space Telescope (JWST) is expected to observe galaxies at z > 10 that are presently inaccessible. Here, we use a self-consistent empirical model, the universemachine, to ...generate mock galaxy catalogues and light-cones over the redshift range z = 0−15. These data include realistic galaxy properties (stellar masses, star formation rates, and UV luminosities), galaxy–halo relationships, and galaxy–galaxy clustering. Mock observables are also provided for different model parameters spanning observational uncertainties at z < 10. We predict that Cycle 1 JWST surveys will very likely detect galaxies with M* > 107 M⊙ and/or M1500 < −17 out to at least z ∼ 13.5. Number density uncertainties at z > 12 expand dramatically, so efforts to detect z > 12 galaxies will provide the most valuable constraints on galaxy formation models. The faint-end slopes of the stellar mass/luminosity functions at a given mass/luminosity threshold steepen as redshift increases. This is because observable galaxies are hosted by haloes in the exponentially falling regime of the halo mass function at high redshifts. Hence, these faint-end slopes are robustly predicted to become shallower below current observable limits (M* < 107 M⊙ or M1500 > −17). For reionization models, extrapolating luminosity functions with a constant faint-end slope from M1500 = −17 down to M1500 = −12 gives the most reasonable upper limit for the total UV luminosity and cosmic star formation rate up to z ∼ 12. We compare to three other empirical models and one semi-analytic model, showing that the range of predicted observables from our approach encompasses predictions from other techniques. Public catalogues and light-cones for common fields are available online.
We employ cosmological hydrodynamical simulations to investigate the effects of AGN feedback on the formation of massive galaxies
with present-day stellar masses of
$M_{\rm stel}= 8.8 \times ...10^{10}{\rm -}6.0 \times 10^{11} {\thinspace {\rm M}_{{\odot }}}$
. Using smoothed particle hydrodynamics simulations with a pressure-entropy formulation that allows an improved treatment of contact discontinuities and fluid mixing, we run three sets of simulations of 20 haloes with different AGN feedback models: (1) no feedback, (2) thermal feedback, and (3) mechanical and radiation feedback. We assume that seed black holes are present at early cosmic epochs at the centre of emerging dark matter haloes and trace their mass growth via gas accretion and mergers with other black holes. Both feedback models successfully recover the observed M
BH–σ relation and black hole-to-stellar mass ratio for simulated central early-type galaxies. The baryonic conversion efficiencies are reduced by a factor of 2 compared to models without any AGN feedback at all halo masses. However, massive galaxies simulated with thermal AGN feedback show a factor of ∼10–100 higher X-ray luminosities than observed. The mechanical/radiation feedback model reproduces the observed correlation between X-ray luminosities and velocity dispersion, e.g. for galaxies with σ = 200 km s− 1, the X-ray luminosity is reduced from 1042 erg s− 1 to 1040 erg s− 1. It also efficiently suppresses late-time star formation, reducing the specific star formation rate from 10−10.5 yr− 1 to 10−14 yr− 1 on average and resulting in quiescent galaxies since z = 2, whereas the thermal feedback model shows higher late-time in situ star formation rates than observed.
ABSTRACT
Different properties of dark matter haloes, including growth rate, concentration, interaction history, and spin, correlate with environment in unique, scale-dependent ways. While these halo ...properties are not directly observable, galaxies will inherit their host haloes’ correlations with environment. In this paper, we show how these characteristic environmental signatures allow using measurements of galaxy environment to constrain which dark matter halo properties are most tightly connected to observable galaxy properties. We show that different halo properties beyond mass imprint distinct scale-dependent signatures in both the galaxy two-point correlation function and the distribution of distances to galaxies’ kth nearest neighbours, with features strong enough to be accessible even with low-resolution (e.g. grism) spectroscopy at higher redshifts. As an application, we compute observed two-point correlation functions for galaxies binned by half-mass radius at $z$ = 0 from the Sloan Digital Sky Survey, showing that classic galaxy size models (i.e. galaxy size being proportional to halo spin) as well as other recent proposals show significant tensions with observational data. We show that the agreement with observed clustering can be improved with a simple empirical model in which galaxy size correlates with halo growth.
ABSTRACT
We investigate the build-up of the galactic dynamo and subsequently the origin of a magnetic driven outflow. We use a set-up of an isolated disc galaxy with a realistic circum-galactic ...medium (CGM). We find good agreement of the galactic dynamo with theoretical and observational predictions from the radial and toroidal components of the magnetic field as function of radius and disc scale height. We find several field reversals indicating dipole structure at early times and quadrupole structure at late times. Together with the magnetic pitch angle and the dynamo control parameters Rα, Rω, and D, we present strong evidence for an α2–Ω dynamo. The formation of a bar in the centre leads to further amplification of the magnetic field via adiabatic compression which subsequently drives an outflow. Due to the Parker instability the magnetic field lines rise to the edge of the disc, break out, and expand freely in the CGM driven by the magnetic pressure. Finally, we investigate the correlation between magnetic field and star formation rate. Globally, we find that the magnetic field is increasing as function of the star formation rate surface density with a slope between 0.3 and 0.45 in good agreement with predictions from theory and observations. Locally, we find that the magnetic field can decrease while star formation increases. We find that this effect is correlated with the diffusion of magnetic field from the spiral arms to the interarm regions which we explicitly include by solving the induction equation and accounting for non-linear terms.
ABSTRACT
We present the novel wide and 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 minimized with particle swarm optimization. GalaxyNet reproduces the observed data very accurately and predicts a stellar-to-halo mass relation with a lower normalization 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 normalization of the instantaneous conversion efficiency increases with redshift, but stays constant above z ≳ 0.5. Finally, we use GalaxyNet to populate a cosmic volume of (5.9 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.