scholarly journals Measuring the spectral index of turbulent gas with deep learning from projected density maps

2020 ◽  
Vol 498 (4) ◽  
pp. 5798-5803
Author(s):  
Piero Trevisan ◽  
Mario Pasquato ◽  
Alessandro Ballone ◽  
Michela Mapelli
Keyword(s):  
Deep Learning ◽  
Star Formation ◽  
Spectral Index ◽  
Adaptive Mesh Refinement ◽  
Mean Squared Error ◽  
Initial Conditions ◽  
Star Cluster ◽  
Large Set ◽  
Simulation Code ◽  
Density Maps

ABSTRACT Turbulence plays a key role in star formation in molecular clouds, affecting star cluster primordial properties. As modelling present-day objects hinges on our understanding of their initial conditions, better constraints on turbulence can result in windfalls in Galactic archaeology, star cluster dynamics, and star formation. Observationally, constraining the spectral index of turbulent gas usually involves computing spectra from velocity maps. Here, we suggest that information on the spectral index might be directly inferred from column density maps (possibly obtained by dust emission/absorption) through deep learning. We generate mock density maps from a large set of adaptive mesh refinement turbulent gas simulations using the hydro-simulation code ramses. We train a convolutional neural network (CNN) on the resulting images to predict the turbulence index, optimize hyperparameters in validation and test on a holdout set. Our adopted CNN model achieves a mean squared error of 0.024 in its predictions on our holdout set, over underlying spectral indexes ranging from 3 to 4.5. We also perform robustness tests by applying our model to altered holdout set images, and to images obtained by running simulations at different resolutions. This preliminary result on simulated density maps encourages further developments on real data, where observational biases and other issues need to be taken into account.

scholarly journals EDGE: a new approach to suppressing numerical diffusion in adaptive mesh simulations of galaxy formation

2020 ◽  
Vol 501 (2) ◽  
pp. 1755-1765
Author(s):  
Andrew Pontzen ◽  
Martin P Rey ◽  
Corentin Cadiou ◽  
Oscar Agertz ◽  
Romain Teyssier ◽  
...  
Keyword(s):  
Star Formation ◽  
Galaxy Formation ◽  
Adaptive Mesh Refinement ◽  
Large Scale ◽  
Initial Conditions ◽  
Dwarf Galaxy ◽  
High Redshift ◽  
Morphological Structure ◽  
Adaptive Mesh ◽  
Numerical Diffusion

ABSTRACT We introduce a new method to mitigate numerical diffusion in adaptive mesh refinement (AMR) simulations of cosmological galaxy formation, and study its impact on a simulated dwarf galaxy as part of the ‘EDGE’ project. The target galaxy has a maximum circular velocity of $21\, \mathrm{km}\, \mathrm{s}^{-1}$ but evolves in a region that is moving at up to $90\, \mathrm{km}\, \mathrm{s}^{-1}$ relative to the hydrodynamic grid. In the absence of any mitigation, diffusion softens the filaments feeding our galaxy. As a result, gas is unphysically held in the circumgalactic medium around the galaxy for $320\, \mathrm{Myr}$, delaying the onset of star formation until cooling and collapse eventually triggers an initial starburst at z = 9. Using genetic modification, we produce ‘velocity-zeroed’ initial conditions in which the grid-relative streaming is strongly suppressed; by design, the change does not significantly modify the large-scale structure or dark matter accretion history. The resulting simulation recovers a more physical, gradual onset of star formation starting at z = 17. While the final stellar masses are nearly consistent ($4.8 \times 10^6\, \mathrm{M}_{\odot }$ and $4.4\times 10^6\, \mathrm{M}_{\odot }$ for unmodified and velocity-zeroed, respectively), the dynamical and morphological structure of the z = 0 dwarf galaxies are markedly different due to the contrasting histories. Our approach to diffusion suppression is suitable for any AMR zoom cosmological galaxy formation simulations, and is especially recommended for those of small galaxies at high redshift.

scholarly journals Calibration of a star formation and feedback model for cosmological simulations with enzo

2020 ◽  
Vol 497 (4) ◽  
pp. 5203-5219
Author(s):  
Boon Kiat Oh ◽  
Britton D Smith ◽  
John A Peacock ◽  
Sadegh Khochfar
Keyword(s):  
Star Formation ◽  
Initial Conditions ◽  
Simulation Code ◽  
Cosmological Simulations ◽  
Feedback Parameter ◽  
Stellar Feedback ◽  
Feedback Model ◽  
Wide Range ◽  
Halo Masses ◽  
Parameter Values

ABSTRACT We introduce a new methodology for efficiently tuning sub-grid models of star formation and supernovae feedback in cosmological simulations and at the same time understanding their physical implications. Based on a set of 71 zoom simulations of a Milky Way (MW)-sized halo, we explore the feasibility of calibrating a widely used star formation and feedback model in the enzo simulation code. We propose a novel way to match observations, using functional fits to the observed baryon makeup over a wide range of halo masses. The model MW galaxy is calibrated using three parameters: the star formation efficiency (f*), the efficiency of thermal energy from stellar feedback (ϵ), and the region into which feedback is injected (r  and s). We find that changing the amount of feedback energy affects the baryon content most significantly. We then identify two sets of feedback parameter values that are both able to reproduce the baryonic properties for haloes between $10^{10}\, \mathrm{M_\odot }$ and $10^{12}\, \mathrm{M_\odot }$. We can potentially improve the agreement by incorporating more parameters or physics. If we choose to focus on one property at a time, we can obtain a more realistic halo baryon makeup. Contrasting both star formation criteria and the corresponding combination of optimal feedback parameters, we also highlight that feedback effects can be complementary: to match the same baryonic properties, with a relatively higher gas-to-stars conversion efficiency, the feedback strength required is lower, and vice versa. Lastly, we demonstrate that chaotic variance in the code can cause deviations of approximately 10 per cent and 25 per cent in the stellar and baryon mass in simulations evolved from identical initial conditions.

Pop III Supernova Feedback on the Formation of the First Galaxies

2019 ◽  
Vol 15 (S341) ◽  
pp. 253-256 ◽  
Author(s):  
Li-Hsin Chen ◽  
Ke-Jung Chen
Keyword(s):  
Dark Matter ◽  
Star Formation ◽  
Adaptive Mesh Refinement ◽  
Initial Conditions ◽  
Adaptive Mesh ◽  
Building Blocks ◽  
Dark Matter Halos ◽  
First Stars ◽  
Stellar Feedback ◽  
First Galaxies

AbstractModern cosmological simulations suggest that the hierarchical assembly of dark matter halos provided the gravitational wells that allowed the primordial gases to form stars and galaxies inside them. The first galaxies comprised of the first systems of stars gravitationally bound in dark matter halos are naturally recognized as the building blocks of early Universe. To understand the formation of the first galaxies, we use an adaptive mesh refinement (AMR) cosmological code, Enzo to simulate the formation and evolution of the first galaxies. We first model an isolated galaxy by considering much microphysics such as star formation, stellar feedback, and primordial gas cooling. To examine the effect of Pop III stellar feedback to the first galaxy formation, we adjust the initial temperature, density distribution and metallicity distributions by assuming different IMFs of the first stars. Our results suggest that star formation in the first galaxies is sensitive to the initial conditions of Pop III supernovae and their remnants. Our study can help to correlate the populations of the first stars and supernovae to star formation inside these first galaxies which may be soon observed by the (James Webb Space Telescope JWST).

scholarly journals Dynamical Evolution of Star Clusters with Intermediate Mass Black Holes and Primordial Binaries

2007 ◽  
Vol 3 (S246) ◽  
pp. 256-260
Author(s):  
Michele Trenti
Keyword(s):  
Black Hole ◽  
Binary Stars ◽  
Initial Conditions ◽  
Relaxation Times ◽  
Star Clusters ◽  
Dynamical Evolution ◽  
Star Cluster ◽  
Core Collapse ◽  
Large Set ◽  
Intermediate Mass

AbstractThe evolution of a star cluster is strongly influenced by the presence of primordial binaries and of a central black hole, as dynamical interactions within the core prevents a deep core collapse under these conditions. We present the results from a large set of direct N-body simulations of star clusters that include an intermediate mass black hole, single and binary stars. We highlight the structural and dynamical differences for the various cases showing in particular that on a timescale of a few relaxation times the density profile of the star cluster does no longer depend on the details of the initial conditions but only on the efficiency of the energy generation due to gravitational encounters at the center of the system.

scholarly journals Auto-Colorization of Historical Images Using Deep Convolutional Neural Networks

Mathematics ◽  
2020 ◽  
Vol 8 (12) ◽  
pp. 2258
Author(s):  
Madhab Raj Joshi ◽  
Lewis Nkenyereye ◽  
Gyanendra Prasad Joshi ◽  
S. M. Riazul Islam ◽  
Mohammad Abdullah-Al-Wadud ◽  
...  
Keyword(s):  
Neural Network ◽  
Deep Learning ◽  
User Study ◽  
Mean Squared Error ◽  
Color Image ◽  
Machine Learning Techniques ◽  
Global Features ◽  
Black And White ◽  
Historical Images ◽  
Learning Techniques

Enhancement of Cultural Heritage such as historical images is very crucial to safeguard the diversity of cultures. Automated colorization of black and white images has been subject to extensive research through computer vision and machine learning techniques. Our research addresses the problem of generating a plausible colored photograph of ancient, historically black, and white images of Nepal using deep learning techniques without direct human intervention. Motivated by the recent success of deep learning techniques in image processing, a feed-forward, deep Convolutional Neural Network (CNN) in combination with Inception- ResnetV2 is being trained by sets of sample images using back-propagation to recognize the pattern in RGB and grayscale values. The trained neural network is then used to predict two a* and b* chroma channels given grayscale, L channel of test images. CNN vividly colorizes images with the help of the fusion layer accounting for local features as well as global features. Two objective functions, namely, Mean Squared Error (MSE) and Peak Signal-to-Noise Ratio (PSNR), are employed for objective quality assessment between the estimated color image and its ground truth. The model is trained on the dataset created by ourselves with 1.2 K historical images comprised of old and ancient photographs of Nepal, each having 256 × 256 resolution. The loss i.e., MSE, PSNR, and accuracy of the model are found to be 6.08%, 34.65 dB, and 75.23%, respectively. Other than presenting the training results, the public acceptance or subjective validation of the generated images is assessed by means of a user study where the model shows 41.71% of naturalness while evaluating colorization results.

scholarly journals Do ultracompact dwarf galaxies form monolithically or as merged star cluster complexes?

2021 ◽  
Vol 502 (4) ◽  
pp. 5185-5199
Author(s):  
Hamidreza Mahani ◽  
Akram Hasani Zonoozi ◽  
Hosein Haghi ◽  
Tereza Jeřábková ◽  
Pavel Kroupa ◽  
...  
Keyword(s):  
Dwarf Galaxies ◽  
Initial Conditions ◽  
Mass Function ◽  
Star Cluster ◽  
Initial Mass Function ◽  
Central Mass ◽  
Cluster Complexes ◽  
V Band ◽  
Stellar Masses ◽  
The Imf

ABSTRACT Some ultracompact dwarf galaxies (UCDs) have elevated observed dynamical V-band mass-to-light (M/LV) ratios with respect to what is expected from their stellar populations assuming a canonical initial mass function (IMF). Observations have also revealed the presence of a compact dark object in the centres of several UCDs, having a mass of a few to 15 per cent of the present-day stellar mass of the UCD. This central mass concentration has typically been interpreted as a supermassive black hole, but can in principle also be a subcluster of stellar remnants. We explore the following two formation scenarios of UCDs: (i) monolithic collapse and (ii) mergers of star clusters in cluster complexes as are observed in massively starbursting regions. We explore the physical properties of the UCDs at different evolutionary stages assuming different initial stellar masses of the UCDs and the IMF being either universal or changing systematically with metallicity and density according to the integrated Galactic IMF theory. While the observed elevated M/LV ratios of the UCDs cannot be reproduced if the IMF is invariant and universal, the empirically derived IMF that varies systematically with density and metallicity shows agreement with the observations. Incorporating the UCD-mass-dependent retention fraction of dark remnants improves this agreement. In addition, we apply the results of N-body simulations to young UCDs and show that the same initial conditions describing the observed M/LV ratios reproduce the observed relation between the half-mass radii and the present-day masses of the UCDs. The findings thus suggest that the majority of UCDs that have elevated M/LV ratios could have formed monolithically with significant remnant-mass components that are centrally concentrated, while those with small M/LV values may be merged star cluster complexes.

scholarly journals A Citizen Science Unmanned Aerial System Data Acquisition Protocol and Deep Learning Techniques for the Automatic Detection and Mapping of Marine Litter Concentrations in the Coastal Zone

Drones ◽  
2021 ◽  
Vol 5 (1) ◽  
pp. 6
Author(s):  
Apostolos Papakonstantinou ◽  
Marios Batsaris ◽  
Spyros Spondylidis ◽  
Konstantinos Topouzelis
Keyword(s):  
Deep Learning ◽  
Data Acquisition ◽  
Citizen Science ◽  
Coastal Zone ◽  
Automatic Detection ◽  
Monitoring Methods ◽  
Monitoring Method ◽  
Marine Litter ◽  
Density Maps ◽  
Learning Techniques

Marine litter (ML) accumulation in the coastal zone has been recognized as a major problem in our time, as it can dramatically affect the environment, marine ecosystems, and coastal communities. Existing monitoring methods fail to respond to the spatiotemporal changes and dynamics of ML concentrations. Recent works showed that unmanned aerial systems (UAS), along with computer vision methods, provide a feasible alternative for ML monitoring. In this context, we proposed a citizen science UAS data acquisition and annotation protocol combined with deep learning techniques for the automatic detection and mapping of ML concentrations in the coastal zone. Five convolutional neural networks (CNNs) were trained to classify UAS image tiles into two classes: (a) litter and (b) no litter. Testing the CCNs’ generalization ability to an unseen dataset, we found that the VVG19 CNN returned an overall accuracy of 77.6% and an f-score of 77.42%. ML density maps were created using the automated classification results. They were compared with those produced by a manual screening classification proving our approach’s geographical transferability to new and unknown beaches. Although ML recognition is still a challenging task, this study provides evidence about the feasibility of using a citizen science UAS-based monitoring method in combination with deep learning techniques for the quantification of the ML load in the coastal zone using density maps.

scholarly journals Importance of the initial conditions for star formation - III. Statistical properties of embedded protostellar clusters

2012 ◽  
Vol 420 (4) ◽  
pp. 3264-3280 ◽  
Author(s):  
Philipp Girichidis ◽  
Christoph Federrath ◽  
Richard Allison ◽  
Robi Banerjee ◽  
Ralf S. Klessen
Keyword(s):  
Star Formation ◽  
Initial Conditions ◽  
Statistical Properties

scholarly journals Turbulent structure and star formation in a stratified, supernova-driven, interstellar medium

2006 ◽  
Vol 2 (S237) ◽  
pp. 358-362
Author(s):  
M. K. Ryan Joung ◽  
Mordecai-Mark Mac Low
Keyword(s):  
Star Formation ◽  
Interstellar Medium ◽  
Adaptive Mesh Refinement ◽  
Formation Rate ◽  
Three Dimensional ◽  
Adaptive Mesh ◽  
Mass Function ◽  
Initial Mass Function ◽  
Interstellar Turbulence ◽  
Self Gravity

AbstractWe report on a study of interstellar turbulence driven by both correlated and isolated supernova explosions. We use three-dimensional hydrodynamic models of a vertically stratified interstellar medium run with the adaptive mesh refinement code Flash at a maximum resolution of 2 pc, with a grid size of 0.5 × 0.5 × 10 kpc. Cold dense clouds form even in the absence of self-gravity due to the collective action of thermal instability and supersonic turbulence. Studying these clouds, we show that it can be misleading to predict physical properties such as the star formation rate or the stellar initial mass function using numerical simulations that do not include self-gravity of the gas. Even if all the gas in turbulently Jeans unstable regions in our simulation is assumed to collapse and form stars in local freefall times, the resulting total collapse rate is significantly lower than the value consistent with the input supernova rate. The amount of mass available for collapse depends on scale, suggesting a simple translation from the density PDF to the stellar IMF may be questionable. Even though the supernova-driven turbulence does produce compressed clouds, it also opposes global collapse. The net effect of supernova-driven turbulence is to inhibit star formation globally by decreasing the amount of mass unstable to gravitational collapse.

Sign in / Sign up

Export Citation Format

Share Document