arXiv subjects

Michael McCourt

Publications and source records attributed to Michael McCourt.

At least 37 records · Page 2Linked to original sources

arXiv2016

Preemptive Termination of Suggestions during Sequential Kriging Optimization of a Brain Activity Reconstruction Simulation

Reconstructing brain activity through electroencephalography requires a boundary value problem (BVP) solver to take a proposed distribution of current dipoles within the brain and compute the resulting electrostatic potential on the scalp. This article proposes the use of sequential kriging optimization to identify different optimal BVP solver parameters for dipoles located in isolated sections of the brain by considering the cumulative impact of randomly oriented dipoles within a chosen isolated section. We attempt preemptive termination of parametrizations suggested during the sequential kriging optimization which, given the results to that point, seem unlikely to produce high quality solutions. Numerical experiments on a simplification of the full geometry for which an approximate solution is available show a benefit from this preemptive termination.

cs.CE↗

arXiv2016

Bayesian Optimization for Machine Learning : A Practical Guidebook

The engineering of machine learning systems is still a nascent field; relying on a seemingly daunting collection of quickly evolving tools and best practices. It is our hope that this guidebook will serve as a useful resource for machine learning practitioners looking to take advantage of Bayesian optimization techniques. We outline four example machine learning problems that can be solved using open source machine learning libraries, and highlight the benefits of using Bayesian optimization in the context of these common machine learning applications.

cs.LG↗

arXiv2016

From Mirrors to Windows: Lyman-Alpha Radiative Transfer in a Very Clumpy Medium

Lyman-Alpha (Ly$α$) is the strongest emission line in the Universe and is frequently used to detect and study the most distant galaxies. Because Lya is a resonant line, photons typically scatter prior to escaping; this scattering process complicates the interpretation of Ly$α$ spectra, but also encodes a wealth of information about the structure and kinematics of neutral gas in the galaxy. Modeling the Ly$α$ line therefore allows us to study tiny-scale features of the gas, even in the most distant galaxies. Curiously, observed Ly$α$ spectra can be modeled successfully with very simple, homogeneous geometries (such as an expanding, spherical shell), whereas more realistic, multiphase geometries often fail to reproduce the observed spectra. This seems paradoxical since the gas in galaxies is known to be multiphase. In this Letter, we show that spectra emerging from extremely clumpy geometries with many clouds along the line of sight converge to the predictions from simplified, homogeneous models. We suggest that this resolves the apparent discrepancy, and may provide a way to study the gas structure in galaxies on scales far smaller than can be probed in either cosmological simulations or direct (i.e., spatially-resolved) observations.

astro-ph.GA↗

arXiv2016

A Characteristic Scale for Cold Gas

We find that clouds of optically-thin, pressure-confined gas are prone to fragmentation as they cool below $\sim10^6$ K. This fragmentation follows the lengthscale $\sim{c}_{\text{s}}\,t_{\text{cool}}$, ultimately reaching very small scales ($\sim{0.1} \text{pc}/n$) as they reach the temperature $\sim10^4$ K at which hydrogen recombines. While this lengthscale depends on the ambient pressure confining the clouds, we find that the column density through an individual fragment $N_{\text{cloudlet}}\sim10^{17} \text{cm}^{-3}$ is essentially independent of environment; this column density represents a characteristic scale for atomic gas at $10^4$ K. We therefore suggest that "clouds" of cold, atomic gas may in fact have the structure of a mist or a fog, composed of tiny fragments dispersed throughout the ambient medium. We show that this scale emerges in hydrodynamic simulations, and that the corresponding increase in the surface area may imply rapid entrainment of cold gas. We also apply it to a number of observational puzzles, including the large covering fraction of diffuse gas in galaxy halos, the broad line widths seen in quasar and AGN spectra, and the entrainment of cold gas in galactic winds. While our simulations make a number of assumptions and thus have associated uncertainties, we show that this characteristic scale is consistent with a number of observations, across a wide range of astrophysical environments. We discuss future steps for testing, improving, and extending our model.

astro-ph.GA↗

arXiv2016

Evaluation System for a Bayesian Optimization Service

Bayesian optimization is an elegant solution to the hyperparameter optimization problem in machine learning. Building a reliable and robust Bayesian optimization service requires careful testing methodology and sound statistical analysis. In this talk we will outline our development of an evaluation framework to rigorously test and measure the impact of changes to the SigOpt optimization service. We present an overview of our evaluation system and discuss how this framework empowers our research engineers to confidently and quickly make changes to our core optimization engine

cs.LG↗

arXiv2016

A Stratified Analysis of Bayesian Optimization Methods

Empirical analysis serves as an important complement to theoretical analysis for studying practical Bayesian optimization. Often empirical insights expose strengths and weaknesses inaccessible to theoretical analysis. We define two metrics for comparing the performance of Bayesian optimization methods and propose a ranking mechanism for summarizing performance within various genres or strata of test functions. These test functions serve to mimic the complexity of hyperparameter optimization problems, the most prominent application of Bayesian optimization, but with a closed form which allows for rapid evaluation and more predictable behavior. This offers a flexible and efficient way to investigate functions with specific properties of interest, such as oscillatory behavior or an optimum on the domain boundary.

cs.LG↗

arXiv2016

Using gas clouds to probe the accretion flow around SgrA*: G2's delayed pericenter passage

We study the dynamical evolution of the putative gas clouds G1 and G2 recently discovered in the Galactic center. Following earlier studies suggesting that these two clouds are part of a larger gas streamer, we combine their orbits into a single trajectory. Since the gas clouds experience a drag force from background gas, this trajectory is not exactly Keplerian. By assuming the G1 and G2 clouds trace this trajectory, we fit for the drag force they experience and thus extract information about the accretion flow at a distance of thousands of Schwarzschild radii from the black hole. This range of radii is important for theories of black hole accretion, but is currently unconstrained by observations. In this paper we extend our previous work by accounting for radial forces due to possible inflow or outflow of the background gas. Such radial forces drive precession in the orbital plane, allowing a slightly better fit to the G1 and G2 data. This precession delays the pericenter passage of G2 by 4-5 months relative to estimates derived from a Keplerian orbital fit; if it proves possible to identify the pericenter time observationally, this enables an immediate test of whether G1 and G2 are gas clouds part of a larger gas streamer. If G2 is indeed a gas cloud, its closest approach likely occurred in late summer 2014, after many of the observing campaigns monitoring G2's anticipated pericenter passage ended. We discuss how this affects interpretation of the G2 observations.

astro-ph.HE↗

arXiv2015

A new inclination instability reshapes Keplerian disks into cones: application to the outer Solar System

Disks of bodies orbiting a much more massive central object are extremely common in astrophysics. When the orbits comprising such disks are eccentric, we show they are susceptible to a new dynamical instability. Gravitational forces between bodies in the disk drive exponential growth of their orbital inclinations and clustering in their angles of pericenter, expanding an initially thin disk into a conical shape by giving each orbit an identical 'tilt' with respect to the disk plane. This new instability dynamically produces the unusual distribution of orbits observed for minor planets beyond Neptune, suggesting that the instability has shaped the outer Solar System. It also implies a large initial disk mass (1-10 Earth masses) of scattered bodies at hundreds of AU; we predict increasing numbers of detections of minor planets clustered in their angles of pericenter with high inclinations.

astro-ph.EP↗

arXiv2015

Unbound Debris Streams and Remnants Resulting From the Tidal Disruptions of Stars by Supermassive Black Holes

The kinetic energy of a star in orbit about a supermassive black hole is a significant fraction of its rest mass energy when its periapse is comparable to its tidal radius. Upon its destruction, a fraction of this energy is extracted and injected into the stellar debris, half of which becomes unbound from the black hole, with the fastest material moving at $\sim 0.03 c$. In this paper, we present a formalism for determining the fate of these unbound debris streams (UDSs) as they depart from the black hole and interact with the surrounding gas. As the density and velocity varies along the length of a UDS, we find that hydrodynamical drag quickly shapes UDSs into loop-like structures, with the densest portions of the streams leading portions of lower density. As UDSs travel outwards, their drag against the ISM increases quadratically with distance, which causes UDSs to deposit their momentum and energy into the ambient medium before the surrounding shocked ISM has a chance to cool. This sudden injection of $\sim 10^{50}$ erg into the ambient medium generates a Sedov-like unbound debris remnant (UDR) that mimics supernova remnants (SNRs) in energetics and appearance, accelerates particles which will produce cosmic rays and synchrotron emission, and provides momentum feedback into the molecular clouds surrounding a black hole. We estimate that a few of these UDRs might be present within a couple degrees of the Galactic Center masquerading as SNRs, and that the UDR scenario is a plausible explanation for Sgr A East.

astro-ph.HE↗

arXiv2015

Thermal Conduction and Multiphase Gas in Cluster Cores

We examine the role of thermal conduction and magnetic fields in cores of galaxy clusters through global simulations of the intracluster medium (ICM). In particular, we study the influence of thermal conduction, both isotropic and anisotropic, on the condensation of multiphase gas in cluster cores. Previous hydrodynamic simulations have shown that cold gas condenses out of the hot ICM in thermal balance only when the ratio of the cooling time ($t_{\rm cool}$) and the free-fall time ($t_{\rm ff}$) is less than $\approx 10$. Since thermal conduction is significant in the ICM and it suppresses local cooling at small scales, it is imperative to include thermal conduction in such studies. We find that anisotropic (along local magnetic field lines) thermal conduction does not influence the condensation criterion for a general magnetic geometry, even if thermal conductivity is large. However, with isotropic thermal conduction cold gas condenses only if conduction is suppressed (by a factor $\lesssim 0.3$) with respect to the Spitzer value.

astro-ph.CO↗

arXiv2015

Going with the flow: using gas clouds to probe the accretion flow feeding Sgr A*

The massive black hole in our galactic center, Sgr A*, accretes only a small fraction of the gas available at its Bondi radius. The physical processes determining this accretion rate remain unknown, partly due to a lack of observational constraints on the gas at distances between ~10 and ~10$^5$ Schwarzschild radii (Rs) from the black hole. Recent infrared observations identify low-mass gas clouds, G1 and G2, moving on highly eccentric, nearly co-planar orbits through the accretion flow around Sgr A*. Although it is not yet clear whether these objects contain embedded stars, their extended gaseous envelopes evolve independently as gas clouds. In this paper we attempt to use these gas clouds to constrain the properties of the accretion flow at ~10$^3$ Rs. Assuming that G1 and G2 follow the same trajectory, we model the small differences in their orbital parameters as evolution resulting from interaction with the background flow. We find evolution consistent with the G-clouds originating in the clockwise disk. Our analysis enables the first unique determination of the rotation axis of the accretion flow: we localize the rotation axis to within 20 degrees, finding an orientation consistent with the parsec-scale jet identified in x-ray observations and with the circumnuclear disk, a massive torus of molecular gas ~1.5 pc from Sgr A*. This suggests that the gas in the accretion flow comes predominantly from the circumnuclear disk, rather than the winds of stars in the young clockwise disk. This result will be tested by the Event Horizon Telescope within the next year. Our model also makes testable predictions for the orbital evolution of G1 and G2, falsifiable on a 5-10 year timescale.

astro-ph.HE↗

arXiv2014

Magnetized Gas Clouds can Survive Acceleration by a Hot Wind

We present three-dimensional magnetohydrodynamic simulations of magnetized gas clouds accelerated by hot winds. We initialize gas clouds with tangled internal magnetic fields and show that this field suppresses the disruption of the cloud: rather than mixing into the hot wind as found in hydrodynamic simulations, cloud fragments end up co-moving and in pressure equilibrium with their surroundings. We also show that a magnetic field in the hot wind enhances the drag force on the cloud by a factor ~(1+v_A^2/v_wind^2)$, where v_A is the Alfven speed in the wind and v_wind measures the relative speed between the cloud and the wind. We apply this result to gas clouds in several astrophysical contexts, including galaxy clusters, galactic winds, the Galactic center, and the outskirts of the Galactic halo. Our results can explain the prevalence of cool gas in galactic winds and galactic halos and how such cool gas survives in spite of its interaction with hot wind/halo gas. We also predict that drag forces can lead to a deviation from Keplerian orbits for the G2 cloud in the galactic center.

astro-ph.GA↗

arXiv2014

Bent Radio Jets Reveal a Stripped Interstellar Medium in NGC 1272

We report the discovery of bent double jets in the elliptical galaxy NGC 1272, a member of the Perseus cluster. The jets have a radius of curvature of ~2 kpc, much smaller than the galaxy half-light radius of ~11 kpc. This bending is likely a result of ram pressure, and indicates that the intracluster gas enters deep within the galaxy and that the interstellar medium of NGC 1272 has been significantly removed. X-ray observations of the surrounding intracluster medium allow us to constrain the pressure within the jet. We find that the standard assumptions of equipartition often used in interpreting other jets underestimate the pressure in the jets by a factor ~30.

astro-ph.GA↗

arXiv2013

What Sets Temperature Gradients in Galaxy Clusters? Implications for non-thermal pressure support and mass-observable scaling relations

We present a spherically symmetric model for the origin and evolution of the temperature profiles in the hot plasma filling galaxy groups and clusters. We find that the gas in clusters is generically not isothermal, and that the temperature declines with radius at large distances from the cluster center (outside the core- and scale radii). This temperature profile is determined by the accretion history of the halo, and is not quantitatively well-described by a polytropic model. We explain quantitatively how the large-scale temperature gradient persists in spite of thermal conduction and convection. These results are a consequence of the cosmological assembly of clusters and cannot be reproduced with non-cosmological simulations of isolated halos. We show that the variation in halo assembly histories produces a ~10% scatter in temperature at fixed mass. On top of this scatter, conduction decreases the temperature of the gas near the scale radius in massive clusters, which may bias hydrostatic mass estimates inferred from x-ray and SZ observations. As an example application of our model profiles, we use mixing-length theory to estimate the turbulent pressure support created by the magnetothermal instability (MTI): in agreement with our earlier MHD simulations, we find that the convection produced by the MTI can provide ~5% non-thermal pressure support near r_500. The magnitude of this turbulent pressure support is likely to be non-monotonic in halo mass, peaking in ~10^14.5 M_sun halos.

astro-ph.CO↗

arXiv2012

Thermal Instability in Gravitationally-Stratified Plasmas: Implications for Multi-Phase Structure in Clusters and Galaxy Halos

We study the interplay among cooling, heating, conduction, and magnetic fields in gravitationally stratified plasmas using simplified, plane-parallel numerical simulations. Since the physical heating mechanism remains uncertain in massive halos such as groups or clusters, we adopt a simple, observationally-motivated prescription which enforces global thermal equilibrium when averaged over large scales. The plasma remains susceptible to local thermal instability, however, and cooling drives an inward flow of material. In contrast to previous results, we argue that the thermal stability of the plasma is independent of its convective stability. We find that the ratio of the cooling timescale to the dynamical timescale t_cool/t_ff controls the saturation of the thermal instability: when t_cool/t_ff < 1, the plasma develops extended multi-phase structure, whereas when t_cool / t_ff > 1 it does not. (In a companion paper, we show that the criterion for thermal instability in a spherical potential is somewhat less stringent, t_cool / t_ff < 10.) When thermal conduction is anisotropic with respect to the magnetic field, the criterion for multi-phase structure is essentially independent of the thermal conductivity of the plasma. Our criterion for local thermal instability to produce multi-phase structure is an extension of the cold vs. hot accretion modes in galaxy formation that applies at all radii in hot halos, not just to the virial shock. We show that this criterion is consistent with data on multi-phase gas in the ACCEPT sample of clusters; in addition, when t_cool / t_ff > 1, the net cooling rate to low temperatures and the mass flux to small radii are suppressed enough relative to models without heating to be qualitatively consistent with star formation rates and x-ray line emission in groups and clusters.

astro-ph.CO↗

arXiv2012

On the Structure of Hot Gas in Halos: Implications for the Lx-Tx Relation & Missing Baryons

We present one-dimensional models of the hot gas in dark-matter halos, which both predict the existence of cool cores and explain their structure. Our models are directly applicable to semi-analytic models (SAMs) of galaxy formation. We have previously argued that filaments of cold (~10^4 K) gas condense out of the intracluster medium (ICM) in hydrostatic and thermal equilibrium when the ratio of the thermal instability timescale to the free-fall time $t_{TI}/t_{ff}$ falls below 5-10. This criterion corresponds to an upper limit on the density of the ICM and motivates a model in which a density core forms wherever $t_{TI}/t_{ff} \lesssim 10$. Consistent with observations and numerical simulations, this model predicts larger and more tenuous cores for lower-mass halos---while the core density in a cluster may be as large as ~ 0.1 cm^{-3}, the core density in the Galactic halo should not exceed ~ 10^{-4} cm^{-3}. We can also explain the large densities in smaller mass halos (galactic `coronae') if we include the contribution of the central galaxy to the gravitational potential. Our models produce a favorable match to the observational X-ray luminosity-temperature (Lx-Tx) relation. For halo masses $\lesssim 10^{13}$ Msun the core size approaches the virial radius. Thus, most of the baryons in such halos cannot be in the hot ICM, but either in the form of stars or in the form of hot gas beyond the virial radius. Because of the smaller mass in the ICM and much larger mass available for star formation, the majority of the baryons in low mass halos ($\lesssim 10^{13}$ Msun) can be expelled beyond the virial radius due to supernova feedback. This can account for the baryons `missing' from low mass halos, such as the Galactic halo.

astro-ph.GA↗

arXiv2012

The Effects of Anisotropic Viscosity on Turbulence and Heat Transport in the Intracluster Medium

In the intracluster medium (ICM) of galaxy clusters, heat and momentum are transported almost entirely along (but not across) magnetic field lines. We perform the first fully self-consistent Braginskii-MHD simulations of galaxy clusters including both of these effects. Specifically, we perform local and global simulations of the magnetothermal instability (MTI) and the heat-flux-driven buoyancy instability (HBI) and assess the effects of viscosity on their saturation and astrophysical implications. We find that viscosity has only a modest effect on the saturation of the MTI. As in previous calculations, we find that the MTI can generate nearly sonic turbulent velocities in the outer parts of galaxy clusters, although viscosity somewhat suppresses the magnetic field amplification. At smaller radii in cool-core clusters, viscosity can decrease the linear growth rates of the HBI. However, it has less of an effect on the HBI's nonlinear saturation, in part because three-dimensional interchange motions (magnetic flux tubes slipping past each other) are not damped by anisotropic viscosity. In global simulations of cool core clusters, we show that the HBI robustly inhibits radial thermal conduction and thus precipitates a cooling catastrophe. The effects of viscosity are, however, more important for higher entropy clusters. We argue that viscosity can contribute to the global transition of cluster cores from cool-core to non cool-core states: additional sources of intracluster turbulence, such as can be produced by AGN feedback or galactic wakes, suppress the HBI, heating the cluster core by thermal conduction; this makes the ICM more viscous, which slows the growth of the HBI, allowing further conductive heating of the cluster core and a transition to a non cool-core state.

astro-ph.CO↗

arXiv2011

Turbulent Pressure Support in the Outer Parts of Galaxy Clusters

We use three-dimensional MHD simulations with anisotropic thermal conduction to study turbulence due to the magnetothermal instability (MTI) in the intracluster medium (ICM) of galaxy clusters. The MTI grows on timescales of ~1 Gyr and is capable of driving vigorous, sustained turbulence in the outer parts of galaxy clusters if the temperature gradient is maintained in spite of the rapid thermal conduction. If this is the case, turbulence due to the MTI can provide up to 5-30% of the pressure support beyond r_500 in galaxy clusters, an effect that is strongest for hot, massive clusters. The turbulence driven by the MTI is generally additive to other sources of turbulence in the ICM, such as that produced by structure formation. This new source of non-thermal pressure support reduces the observed Sunyaev-Zel'dovich (SZ) signal and X-ray pressure gradient for a given cluster mass and introduces a cluster mass and temperature gradient-dependent bias in SZ and X-ray mass estimates of clusters. This additional physics may also need to be taken into account when estimating the matter power spectrum normalization, sigma-8, through simulation templates from the observed amplitude of the SZ power spectrum.

astro-ph.CO↗