Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “catastrophe”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 739 records · Page 41

Mean and extreme radio properties of quasars and the origin of radio emission

We investigate the evolution of both the radio-loud fraction (RLF) and (using stacking analysis) the mean radio loudness of quasars. We consider how these properties evolve as a function of redshift and luminosity, black hole (BH) mass and accretion rate, and parameters related to the dominance of a wind in the broad emission-line region. We match the FIRST source catalog to samples of luminous quasars (both spectroscopic and photometric), primarily from the Sloan Digital Sky Survey. After accounting for catastrophic errors in BH mass estimates at high redshift, we find that both the RLF and the mean radio luminosity increase for increasing BH mass and decreasing accretion rate. Similarly, both the RLF and mean radio loudness increase for quasars that are argued to have weaker radiation line driven wind components of the broad emission-line region. In agreement with past work, we find that the RLF increases with increasing optical/UV luminosity and decreasing redshift, while the mean radio loudness evolves in the exact opposite manner. This difference in behavior between the mean radio loudness and the RLF in L−z may indicate selection effects that bias our understanding of the evolution of the RLF; deeper surveys in the optical and radio are needed to resolve this discrepancy. Finally, we argue that radio-loud (RL) and radio-quiet (RQ) quasars may be parallel sequences, but where only RQ quasars at one extreme of the distribution are likely to become RL, possibly through slight differences in spin and/or merger history.

79 ASTRONOMY AND ASTROPHYSICS↗

A new merging double degenerate binary in the solar neighborhood

Characterizing the local space density of double degenerate (DD) binary systems is a complementary approach to broad sky surveys of DDs to determine the expected rates of WD binary mergers, in particular those that may evolve into other observable phenomena such as extreme helium stars, Am CVn systems, and SNe Ia. However, there have been few such systems detected in local space. We report here the discovery that WD 1242–105, a nearby bright WD, is a double-line spectroscopic binary consisting of two degenerate DA WDs of similar mass and temperature, despite it previously having been spectroscopically characterized as a single degenerate. Follow-up photometry, spectroscopy, and trigonometric parallax have been obtained in an effort to determine the fundamental parameters of each component of this system. The binary has a mass ratio of 0.7 and a trigonometric parallax of 25.5 mas, placing it at a distance of 39 pc. The system’s total mass is 0.95 M{sub ⊙} and has an orbital period of 2.85 hr, making it the strongest known gravitational wave source (logh=−20.78) in the mHz regime. Because of its orbital period and total mass, WD 1242–105 is predicted to merge via gravitational radiation on a timescale of 740 Myr, which will most likely not result in a catastrophic explosion.

79 ASTRONOMY AND ASTROPHYSICS↗

ELECTRIC CHARGING OF DUST AGGREGATES AND ITS EFFECT ON DUST COAGULATION IN PROTOPLANETARY DISKS

Mutual sticking of dust aggregates is the first step toward planetesimal formation in protoplanetary disks. In spite that the electric charging of dust particles is well recognized in some contexts, it has been largely ignored in the current modeling of dust coagulation. In this study, we present a general analysis of the dust charge state in protoplanetary disks, and then demonstrate how the electric charging could dramatically change the currently accepted scenario of dust coagulation. First, we describe a new semianalytical method to calculate the dust charge state and gas ionization state self-consistently. This method is far more efficient than previous numerical methods, and provides a general and clear description of the charge state of a gas-dust mixture. Second, we apply this analysis to compute the collisional cross section of growing aggregates taking their charging into account. As an illustrative example, we focus on early evolutionary stages where the dust has been thought to grow into fractal (D {approx} 2) aggregates with a quasi-monodisperse (i.e., narrow) size distribution. We find that, for a wide range of model parameters, the fractal growth is strongly inhibited by the electric repulsion between colliding aggregates and eventually 'freezes out' on its way to the subsequent growth stage involving collisional compression. Strong disk turbulence would help the aggregates to overcome this growth barrier, but then it would cause catastrophic collisional fragmentation in later growth stages. These facts suggest that the combination of electric repulsion and collisional fragmentation would impose a serious limitation on dust growth in protoplanetary disks. We propose a possible scenario of dust evolution after the freezeout. Finally, we point out that the fractal growth of dust aggregates tends to maintain a low ionization degree and, as a result, a large magnetorotationally stable region in the disk.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

COLLISIONAL STRIPPING AND DISRUPTION OF SUPER-EARTHS

The final stage of planet formation is dominated by collisions between planetary embryos. The dynamics of this stage determine the orbital configuration and the mass and composition of planets in the system. In the solar system, late giant impacts have been proposed for Mercury, Earth, Mars, and Pluto. In the case of Mercury, this giant impact may have significantly altered the bulk composition of the planet. Here we present the results of smoothed particle hydrodynamics simulations of high-velocity (up to {approx}5v {sub esc}) collisions between 1 and 10 M {sub +} planets of initially terrestrial composition to investigate the end stages of formation of extrasolar super-Earths. As found in previous simulations of collisions between smaller bodies, when collision energies exceed simple merging, giant impacts are divided into two regimes: (1) disruption and (2) hit-and-run (a grazing inelastic collision and projectile escape). Disruption occurs when the impact parameter is near zero, when the projectile mass is small compared to the target, or at extremely high velocities. In the disruption regime, we derive the criteria for catastrophic disruption (when half the total colliding mass is lost), the transition energy between accretion and erosion, and a scaling law for the change in bulk composition (iron-to-silicate ratio) resulting from collisional stripping of a mantle.

99 GENERAL AND MISCELLANEOUS↗

SIMULATIONS OF MAGNETOHYDRODYNAMICS INSTABILITIES IN INTRACLUSTER MEDIUM INCLUDING ANISOTROPIC THERMAL CONDUCTION

We perform a suite of simulations of cooling cores in clusters of galaxies in order to investigate the effect of the recently discovered heat flux buoyancy instability (HBI) on the evolution of cores. Our models follow the three-dimensional magnetohydrodynamics of cooling cluster cores and capture the effects of anisotropic heat conduction along the lines of magnetic field, but do not account for the cosmological setting of clusters or the presence of active galactic nuclei (AGNs). Our model clusters can be divided into three groups according to their final thermodynamical state: catastrophically collapsing cores, isothermal cores, and an intermediate group whose final state is determined by the initial configuration of magnetic field. Modeled cores that are reminiscent of real cluster cores show evolution toward thermal collapse on a timescale which is prolonged by a factor of approx2-10 compared with the zero-conduction cases. The principal effect of the HBI is to re-orient field lines to be perpendicular to the temperature gradient. Once the field has been wrapped up onto spherical surfaces surrounding the core, the core is insulated from further conductive heating (with the effective thermal conduction suppressed to less than 10{sup -2} of the Spitzer value) and proceeds to collapse. We speculate that, in real clusters, the central AGN and possibly mergers play the role of 'stirrers', periodically disrupting the azimuthal field structure and allowing thermal conduction to sporadically heat the core.

79 ASTRONOMY AND ASTROPHYSICS↗

SIGNATURES OF IMPULSIVE LOCALIZED HEATING IN THE TEMPERATURE DISTRIBUTION OF MULTI-STRANDED CORONAL LOOPS

We study the signatures of coronal heating on the differential emission measure (DEM) by means of hydrodynamic simulations capable of resolving the chromospheric-corona transition region sections of multi-stranded coronal loops and following their evolution. We consider heating either uniformly distributed along the loop or localized close to the chromospheric footpoints, in both steady and impulsive regimes. Our simulations show that condensation at the top of the loop forms when the impulsive heating, with a pulse cadence lower than the plasma cooling time, is localized at the loop footpoints, and the pulse energy is below a threshold above which the heating balances the radiative losses, thus preventing the catastrophic cooling which triggers the condensation. A condensation does not produce observable signatures in the DEM because it does not redistribute the plasma over a sufficiently large temperature range. On the other hand, the DEM coronal peak is found sensitive to the pulse cadence time when this is longer or comparable to the plasma cooling time. In this case, the heating pulses produce large oscillations in temperature in the bulk of the coronal plasma, which effectively smears out the coronal DEM structure. The pronounced DEM peak observed in active regions would indicate a predominance of conditions in which the cadence time is shorter or of the order of the plasma cooling time, whilst the structure of the quiet-Sun DEM suggests a cadence time longer than the plasma cooling time. Our simulations give an explanation of the warm overdense and hot underdense loops observed by TRACE, SOHO, and Yohkoh. However, they are unable to reproduce both the transition region and the coronal DEM structure with a unique set of parameters, which outlines the need for a more realistic description of the transition region.

79 ASTRONOMY AND ASTROPHYSICS↗

CAN THERMAL NONEQUILIBRIUM EXPLAIN CORONAL LOOPS?

Any successful model of coronal loops must explain a number of observed properties. For warm ({approx}1 MK) loops, these include (1) excess density, (2) flat temperature profile, (3) super-hydrostatic scale height, (4) unstructured intensity profile, and (5) 1000-5000 s lifetime. We examine whether thermal nonequilibrium can reproduce the observations by performing hydrodynamic simulations based on steady coronal heating that decreases exponentially with height. We consider both monolithic and multi-stranded loops. The simulations successfully reproduce certain aspects of the observations, including the excess density, but each of them fails in at least one critical way. Monolithic models have far too much intensity structure, while multi-strand models are either too structured or too long-lived. Our results appear to rule out the widespread existence of heating that is both highly concentrated low in the corona and steady or quasi-steady (slowly varying or impulsive with a rapid cadence). Active regions would have a very different appearance if the dominant heating mechanism had these properties. Thermal nonequilibrium may nonetheless play an important role in prominences and catastrophic cooling events (e.g., coronal rain) that occupy a small fraction of the coronal volume. However, apparent inconsistencies between the models and observations of cooling events have yet to be understood.

79 ASTRONOMY AND ASTROPHYSICS↗

THE FORMATION OF THE COLLISIONAL FAMILY AROUND THE DWARF PLANET HAUMEA

Haumea, a rapidly rotating elongated dwarf planet ({approx}1500 km in diameter), has two satellites and is associated with a 'family' of several smaller Kuiper Belt objects (KBOs) in similar orbits. All members of the Haumea system share a water ice spectral feature that is distinct from all other KBOs. The relative velocities between the Haumea family members are too small to have formed by catastrophic disruption of a large precursor body, which is the process that formed families around much smaller asteroids in the main belt. Here, we show that all of the unusual characteristics of the Haumea system are explained by a novel type of giant collision: a graze and merge impact between two comparably sized bodies. The grazing encounter imparted the high angular momentum that spun off fragments from the icy crust of the elongated merged body. The fragments became satellites and family members. Giant collision outcomes are extremely sensitive to the impact parameters. Compared to the main belt, the largest bodies in the Kuiper Belt are more massive and experience slower velocity collisions; hence, outcomes of giant collisions are dramatically different between the inner and outer solar system. The dwarf planets in the Kuiper Belt record an unexpectedly large number of giant collisions, requiring a special dynamical event at the end of solar system formation.

79 ASTRONOMY AND ASTROPHYSICS↗

CORONAL RAIN AS A MARKER FOR CORONAL HEATING MECHANISMS

Reported observations in H{alpha}, Ca II H, and K or other chromospheric lines of coronal rain trace back to the days of the Skylab mission. Corresponding to cool and dense plasma, coronal rain is often observed falling down along coronal loops in active regions. A physical explanation for this spectacular phenomenon has been put forward thanks to numerical simulations of loops with footpoint-concentrated heating, a heating scenario in which cool condensations naturally form in the corona. This effect has been termed 'catastrophic cooling' and is the predominant explanation for coronal rain. In this work, we further investigate the link between this phenomenon and the heating mechanisms acting in the corona. We start by analyzing observations of coronal rain at the limb in the Ca II H line performed by the Hinode satellite, and derive interesting statistical properties concerning the dynamics. We then compare the observations with 1.5-dimensional MHD simulations of loops being heated by small-scale discrete events concentrated toward the footpoints (that could come, for instance, from magnetic reconnection events), and by Alfven waves generated at the photospheric level. Both our observation and simulation results suggest that coronal rain is a far more common phenomenon than previously thought. Also, we show that the structure and dynamics of condensations are far more sensitive to the internal pressure changes in loops than to gravity. Furthermore, it is found that if a loop is predominantly heated from Alfven waves, coronal rain is inhibited due to the characteristic uniform heating they produce. Hence, coronal rain may not only point to the spatial distribution of the heating in coronal loops but also to the agent of the heating itself. We thus propose coronal rain as a marker for coronal heating mechanisms.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

MASSCLEANage-STELLAR CLUSTER AGES FROM INTEGRATED COLORS

We present the recently updated and expanded MASSCLEANcolors, a database of 70 million Monte Carlo models selected to match the properties (metallicity, ages, and masses) of stellar clusters found in the Large Magellanic Cloud (LMC). This database shows the rather extreme and non-Gaussian distribution of integrated colors and magnitudes expected with different cluster age and mass and the enormous age degeneracy of integrated colors when mass is unknown. This degeneracy could lead to catastrophic failures in estimating age with standard simple stellar population models, particularly if most of the clusters are of intermediate or low mass, like in the LMC. Utilizing the MASSCLEANcolors database, we have developed MASSCLEANage, a statistical inference package which assigns the most likely age and mass (solved simultaneously) to a cluster based only on its integrated broadband photometric properties. Finally, we use MASSCLEANage to derive the age and mass of LMC clusters based on integrated photometry alone. First, we compare our cluster ages against those obtained for the same seven clusters using more accurate integrated spectroscopy. We find improved agreement with the integrated spectroscopy ages over the original photometric ages. A close examination of our results demonstrates the necessity of solving simultaneously for mass and age to reduce degeneracies in the cluster ages derived via integrated colors. We then selected an additional subset of 30 photometric clusters with previously well-constrained ages and independently derive their age using the MASSCLEANage with the same photometry with very good agreement. The MASSCLEANage program is freely available under GNU General Public License.

79 ASTRONOMY AND ASTROPHYSICS↗

ON THE FORMATION OF MULTIPLE STELLAR POPULATIONS IN GLOBULAR CLUSTERS

Nearly all globular clusters (GCs) studied to date show evidence for multiple stellar populations, in stark contrast to the conventional view that GCs are a mono-metallic, coeval population of stars. This generic feature must therefore emerge naturally within massive star cluster formation. Building on earlier work, we propose a simple physical model for the early evolution (several 10{sup 8} yr) of GCs. We consider the effects of stellar mass loss, Type II supernovae (SNe II) and prompt Type Ia supernovae (SNe Ia), ram pressure, and accretion from the ambient interstellar medium (ISM) on the development of a young GC's own gas reservoir. In our model, SNe II from a first generation of star formation clears the GC of its initial gas reservoir. Over the next several 10{sup 8} yr, mass lost from asymptotic giant branch stars and matter accreted from the ambient ISM collect at the center of the GC. This material must remain quite cool (T {approx} 10{sup 2} K), but does not catastrophically cool on a crossing time because of the high Lyman-Werner flux density in young GCs. The collection of gas within the GC must compete with ram pressure from the ambient ISM. After several 10{sup 8} yr, the Lyman-Werner photon flux density drops by more than three orders of magnitude, allowing molecular hydrogen and then stars to form. After this second generation of star formation, SNe II from the second generation and then prompt SNe Ia associated with the first generation maintain a gas-free GC, thereby ending the cycle of star formation events. Our model makes clear predictions for the presence or absence of multiple stellar populations within GCs as a function of GC mass and formation environment. While providing a natural explanation for the approximately equal number of first- and second-generation stars in GCs, substantial accretion from the ambient ISM may produce fewer chemically peculiar second-generation stars than are observed. Analyzing intermediate-age LMC clusters, we find for the first time evidence for a mass threshold of {approx}10{sup 4} M{sub sun} below which LMC clusters appear to be truly coeval. This threshold mass is consistent with our predictions for the mass at which ram pressure is capable of clearing gas from clusters in the LMC at the present epoch. Recently, claims have been made that multiple populations within GCs require that GCs form at the center of their own dark matter halos. We argue that such a scenario is implausible. Observations of the young and intermediate-age clusters in the LMC and M31 will provide strong constraints on our proposed scenario.

79 ASTRONOMY AND ASTROPHYSICS↗

SPIRALING OUT OF CONTROL: THREE-DIMENSIONAL HYDRODYNAMICAL MODELING OF THE COLLIDING WINDS IN {eta} CARINAE

Three-dimensional adaptive mesh refinement hydrodynamical simulations of the wind-wind collision between the enigmatic supermassive star {eta} Car and its mysterious companion star are presented which include radiative driving of the stellar winds, gravity, optically thin radiative cooling, and orbital motion. Simulations with static stars with a periastron passage separation reveal that the preshock companion star's wind speed is sufficiently reduced so that radiative cooling in the postshock gas becomes important, permitting the runaway growth of nonlinear thin-shell instabilities (NTSIs) which massively distort the wind-wind collision region (WCR). However, large-scale simulations, which include the orbital motion of the stars, show that orbital motion reduces the impact of radiative inhibition and thus increases the acquired preshock velocities. As such, the postshock gas temperature and cooling time see a commensurate increase, and sufficient gas pressure is preserved to stabilize the WCR against catastrophic instability growth. We then compute synthetic X-ray spectra and light curves and find that, compared to previous models, the X-ray spectra agree much better with XMM-Newton observations just prior to periastron. The narrow width of the 2009 X-ray minimum can also be reproduced. However, the models fail to reproduce the extended X-ray minimum from previous cycles. We conclude that the key to explaining the extended X-ray minimum is the rate of cooling of the companion star's postshock wind. If cooling is rapid then powerful NTSIs will heavily disrupt the WCR. Radiative inhibition of the companion star's preshock wind, albeit with a stronger radiation-wind coupling than explored in this work, could be an effective trigger.

79 ASTRONOMY AND ASTROPHYSICS↗

THE ABSENCE OF COLD DUST AND THE MINERALOGY AND ORIGIN OF THE WARM DUST ENCIRCLING BD +20 307

Spitzer Space Telescope photometry and spectroscopy of BD +20 307 show that all of the dust around this remarkable Gyr-old spectroscopic binary arises within 1 AU. No additional cold dust is needed to fit the infrared excess. Peaks in the 10 and 20 {mu}m spectrum are well fit with small silicates that should be removed on a timescale of years from the system. This is the dustiest star known for its age, which is {approx}>1 Gyr. The dust cannot arise from a steady-state collisional cascade. A catastrophic collision of two rocky, planetary-scale bodies in the terrestrial zone is the most likely source for this warm dust because it does not require a reservoir of planetesimals in the outer system.

79 ASTRONOMY AND ASTROPHYSICS↗

ARE ALL SHORT-HARD GAMMA-RAY BURSTS PRODUCED FROM MERGERS OF COMPACT STELLAR OBJECTS?

The origin and progenitors of short-hard gamma-ray bursts (GRBs) remain a puzzle and a highly debated topic. Recent Swift observations suggest that these GRBs may be related to catastrophic explosions in degenerate compact stars, denoted as 'Type I' GRBs. The most popular models include the merger of two compact stellar objects (NS-NS or NS-BH). We utilize a Monte Carlo approach to determine whether a merger progenitor model can self-consistently account for all the observations of short-hard GRBs, including a sample with redshift measurements in the Swift era (z-known sample) and the CGRO/BATSE sample. We apply various merger time delay distributions invoked in compact star merger models to derive the redshift distributions of these Type I GRBs, and then constrain the unknown luminosity function of Type I GRBs using the observed luminosity-redshift (L-z) distributions of the z-known sample. The best luminosity function model, together with the adopted merger delay model, is then applied to confront the peak flux distribution (log N-log P distribution) of the BATSE and Swift samples. We find that for all the merger models invoking a range of merger delay timescales (including those invoking a large fraction of 'prompt mergers'), it is difficult to reconcile the models with all the data. The data are instead statistically consistent with the following two possible scenarios. First, that short/hard GRBs are a superposition of compact-star-merger-origin (Type I) GRBs and a population of GRBs that track the star formation history, which are probably related to the deaths of massive stars (Type II GRBs). Second, the entire short/hard GRB population is consistent with a typical delay of 2 Gyr with respect to the star formation history with modest scatter. This may point toward a different Type I progenitor than the traditional compact star merger models.

79 ASTRONOMY AND ASTROPHYSICS↗

OBSERVATIONS AND MAGNETIC FIELD MODELING OF THE FLARE/CORONAL MASS EJECTION EVENT ON 2010 APRIL 8

We present a study of the flare/coronal mass ejection event that occurred in Active Region 11060 on 2010 April 8. This event also involves a filament eruption, EIT wave, and coronal dimming. Prior to the flare onset and filament eruption, both SDO/AIA and STEREO/EUVI observe a nearly horizontal filament ejection along the internal polarity inversion line, where flux cancellations frequently occur as observed by SDO/HMI. Using the flux-rope insertion method developed by van Ballegooijen, we construct a grid of magnetic field models using two magneto-frictional relaxation methods. We find that the poloidal flux is significantly reduced during the relaxation process, though one relaxation method preserves the poloidal flux better than the other. The best-fit pre-flare NLFFF model is constrained by matching the coronal loops observed by SDO/AIA and Hinode/XRT. We find that the axial flux in this model is very close to the threshold of instability. For the model that becomes unstable due to an increase of the axial flux, the reconnected field lines below the X-point closely match the observed highly sheared flare loops at the event onset. The footpoints of the erupting flux rope are located around the coronal dimming regions. Both observational and modeling results support the premise that this event may be initiated by catastrophic loss of equilibrium caused by an increase of the axial flux in the flux rope, which is driven by flux cancellations.

79 ASTRONOMY AND ASTROPHYSICS↗

SLOSHING OF THE MAGNETIZED COOL GAS IN THE CORES OF GALAXY CLUSTERS

X-ray observations of many clusters of galaxies reveal the presence of edges in surface brightness and temperature, known as 'cold fronts'. In relaxed clusters with cool cores, these edges have been interpreted as evidence for the 'sloshing' of the core gas in the cluster's gravitational potential. The smoothness of these edges has been interpreted as evidence for the stabilizing effect of magnetic fields 'draped' around the front surfaces. To check this hypothesis, we perform high-resolution magnetohydrodynamics simulations of magnetized gas sloshing in galaxy clusters initiated by encounters with subclusters. We go beyond previous works on the simulation of cold fronts in a magnetized intracluster medium by simulating their formation in realistic, idealized mergers with high resolution ({Delta}x {approx} 2 kpc). Our simulations sample a parameter space of plausible initial magnetic field strengths and field configurations. In the simulations, we observe strong velocity shears associated with the cold fronts amplifying the magnetic field along the cold front surfaces, increasing the magnetic field strength in these layers by up to an order of magnitude, and boosting the magnetic pressure up to near-equipartition with thermal pressure in some cases. In these layers, the magnetic field becomes strong enough to stabilize the cold fronts against Kelvin-Helmholtz instabilities, resulting in sharp, smooth fronts as those seen in observations of real clusters. These magnetic fields also result in strong suppression of mixing of high- and low-entropy gases in the cluster, seen in our simulations of mergers in the absence of a magnetic field. As a result, the heating of the core due to sloshing is very modest and is unable to stave off a cooling catastrophe.

79 ASTRONOMY AND ASTROPHYSICS↗

THE STEADY-STATE WIND MODEL FOR YOUNG STELLAR CLUSTERS WITH AN EXPONENTIAL STELLAR DENSITY DISTRIBUTION

A hydrodynamic model for steady-state, spherically symmetric winds driven by young stellar clusters with an exponential stellar density distribution is presented. Unlike in most previous calculations, the position of the singular point R{sub sp}, which separates the inner subsonic zone from the outer supersonic flow, is not associated with the star cluster edge, but calculated self-consistently. When the radiative losses of energy are negligible, the transition from the subsonic to the supersonic flow occurs always at R{sub sp} Almost-Equal-To 4R{sub c} , where R{sub c} is the characteristic scale for the stellar density distribution, irrespective of other star cluster parameters. This is not the case in the catastrophic cooling regime, when the temperature drops abruptly at a short distance from the star cluster center, and the transition from the subsonic to the supersonic regime occurs at a much smaller distance from the star cluster center. The impact from the major star cluster parameters to the wind inner structure is thoroughly discussed. Particular attention is paid to the effects which radiative cooling provides to the flow. The results of the calculations for a set of input parameters, which lead to different hydrodynamic regimes, are presented and compared to the results from non-radiative one-dimensional numerical simulations and to those from calculations with a homogeneous stellar mass distribution.

79 ASTRONOMY AND ASTROPHYSICS↗

ACTIVE LEARNING TO OVERCOME SAMPLE SELECTION BIAS: APPLICATION TO PHOTOMETRIC VARIABLE STAR CLASSIFICATION

Despite the great promise of machine-learning algorithms to classify and predict astrophysical parameters for the vast numbers of astrophysical sources and transients observed in large-scale surveys, the peculiarities of the training data often manifest as strongly biased predictions on the data of interest. Typically, training sets are derived from historical surveys of brighter, more nearby objects than those from more extensive, deeper surveys (testing data). This sample selection bias can cause catastrophic errors in predictions on the testing data because (1) standard assumptions for machine-learned model selection procedures break down and (2) dense regions of testing space might be completely devoid of training data. We explore possible remedies to sample selection bias, including importance weighting, co-training, and active learning (AL). We argue that AL-where the data whose inclusion in the training set would most improve predictions on the testing set are queried for manual follow-up-is an effective approach and is appropriate for many astronomical applications. For a variable star classification problem on a well-studied set of stars from Hipparcos and Optical Gravitational Lensing Experiment, AL is the optimal method in terms of error rate on the testing data, beating the off-the-shelf classifier by 3.4% and the other proposed methods by at least 3.0%. To aid with manual labeling of variable stars, we developed a Web interface which allows for easy light curve visualization and querying of external databases. Finally, we apply AL to classify variable stars in the All Sky Automated Survey, finding dramatic improvement in our agreement with the ASAS Catalog of Variable Stars, from 65.5% to 79.5%, and a significant increase in the classifier's average confidence for the testing set, from 14.6% to 42.9%, after a few AL iterations.

79 ASTRONOMY AND ASTROPHYSICS↗