skip to main content

Title: A Comparative Study of Machine-learning Methods for X-Ray Binary Classification

X-ray binaries (XRBs) consist of a compact object that accretes material from an orbiting secondary star. The most secure method we have for determining if the compact object is a black hole is to determine its mass: This is limited to bright objects and requires substantial time-intensive spectroscopic monitoring. With new X-ray sources being discovered with different X-ray observatories, developing efficient, robust means to classify compact objects becomes increasingly important. We compare three machine-learning classification methods (Bayesian Gaussian Processes (BGPs), K-Nearest Neighbors (KNN), Support Vector Machines) for determining whether the compact objects are neutron stars or black holes (BHs) in XRB systems. Each machine-learning method uses spatial patterns that exist between systems of the same type in 3D color–color–intensity diagrams. We used lightcurves extracted using 6 yr of data with MAXI/GSC for 44 representative sources. We find that all three methods are highly accurate in distinguishing pulsing from nonpulsing neutron stars (NPNS) with 95% of NPNS and 100% of pulsars accurately predicted. All three methods have high accuracy in distinguishing BHs from pulsars (92%) but continue to confuse BHs with a subclass of NPNS, called bursters, with KNN doing the best at only 50% accuracy for predicting BHs. The more » precision of all three methods is high, providing equivalent results over 5–10 independent runs. In future work, we will suggest a fourth dimension be incorporated to mitigate the confusion of BHs with bursters. This work paves the way toward more robust methods to efficiently distinguish BHs, NPNS, and pulsars.

« less
; ; ;
Publication Date:
Journal Name:
The Astrophysical Journal
Page Range or eLocation-ID:
Article No. 116
DOI PREFIX: 10.3847
Sponsoring Org:
National Science Foundation
More Like this

    The discs of active galactic nuclei (AGNs) have emerged as rich environments for the production and capture of stars and the compact objects that they leave behind. These stars produce long gamma-ray bursts (GRBs) at their deaths, while frequent interactions among compact objects form binary neutron stars and neutron star–black hole binaries, leading to short GRBs upon their merger. Predicting the properties of these transients as they emerge from the dense environments of AGN discs is key to their proper identification and to better constrain the star and compact object population in AGN discs. Some of these transients would appear unusual because they take place in much higher densities than the interstellar medium. Others, which are the subject of this paper, would additionally be modified by radiation diffusion, since they are generated within optically thick regions of the accretion discs. Here, we compute the GRB afterglow light curves for diffused GRB sources for a representative variety of central black hole masses and disc locations. We find that the radiation from radio to ultraviolet and soft X-rays can be strongly suppressed by synchrotron self-absorption in the dense medium of the AGN disc. In addition, photon diffusion can significantly delay themore »emergence of the emission peak, turning a beamed, fast transient into a slow, isotropic, and dimmer one. These would appear as broad-band correlated AGN variability with a dominance at the higher frequencies. Their properties can constrain both the stellar populations within AGN discs and the disc structure.

    « less
  2. Abstract Reliable neutron star mass measurements are key to determining the equation of state of cold nuclear matter, but such measurements are rare. Black widows and redbacks are compact binaries consisting of millisecond pulsars and semi-degenerate companion stars. Spectroscopy of the optically bright companions can determine their radial velocities, providing inclination-dependent pulsar mass estimates. Although inclinations can be inferred from subtle features in optical light curves, such estimates may be systematically biased due to incomplete heating models and poorly understood variability. Using data from the Fermi Large Area Telescope, we have searched for gamma-ray eclipses from 49 spider systems, discovering significant eclipses in 7 systems, including the prototypical black widow PSR B1957+20. Gamma-ray eclipses require direct occultation of the pulsar by the companion, and so the detection, or significant exclusion, of a gamma-ray eclipse strictly limits the binary inclination angle, providing new robust, model-independent pulsar mass constraints. For PSR B1957+20, the eclipse implies a much lighter pulsar (1.81 ± 0.07 solar masses) than inferred from optical light curve modelling.
  3. Abstract The globular cluster 47 Tucanae (47 Tuc) is one of the most massive star clusters in the Milky Way and is exceptionally rich in exotic stellar populations. For several decades it has been a favorite target of observers, and yet it is computationally very challenging to model because of its large number of stars ( N ≳ 10 6 ) and high density. Here we present detailed and self-consistent 47 Tuc models computed with the Cluster Monte Carlo code ( CMC ). The models include all relevant dynamical interactions coupled to stellar and binary evolution, and reproduce various observations, including the surface brightness and velocity dispersion profiles, pulsar accelerations, and numbers of compact objects. We show that the present properties of 47 Tuc are best reproduced by adopting an initial stellar mass function that is both bottom-heavy and top-light relative to standard assumptions (as in, e.g., Kroupa 2001), and an initial Elson profile (Elson et al. 1987) that is overfilling the cluster’s tidal radius. We include new prescriptions in CMC for the formation of binaries through giant star collisions and tidal captures, and we show that these mechanisms play a crucial role in the formation of neutron star binariesmore »and millisecond pulsars in 47 Tuc; our best-fit model contains ∼50 millisecond pulsars, 70% of which are formed through giant collisions and tidal captures. Our models also suggest that 47 Tuc presently contains up to ∼200 stellar-mass black holes, ∼5 binary black holes, ∼15 low-mass X-ray binaries, and ∼300 cataclysmic variables.« less
  4. Abstract

    The Milky Way is believed to host hundreds of millions of quiescent stellar-mass black holes (BHs). In the last decade, some of these objects have been potentially uncovered via gravitational microlensing events. All these detections resulted in a degeneracy between the velocity and the mass of the lens. This degeneracy has been lifted, for the first time, with the recent astrometric microlensing detection of OB110462. However, two independent studies reported very different lens masses for this event. Sahu et al. inferred a lens mass of 7.1 ± 1.3M, consistent with a BH, while Lam et al. inferred 1.6–4.2M, consistent with either a neutron star or a BH. Here, we study the landscape of isolated BHs formed in the field. In particular, we focus on the mass and center-of-mass speed of four subpopulations: isolated BHs from single-star origin, disrupted BHs of binary-star origin, main-sequence stars with a compact object companion, and double compact object mergers. Our model predicts that most (≳70%) isolated BHs in the Milky Way are of binary origin. However, noninteractions lead to most massive BHs (≳15–20M) being predominantly of single origin. Under the assumption that OB110462 is a free-floating compact object, we conclude that it is moremore »likely to be a BH originally belonging to a binary system. Our results suggest that low-mass BH microlensing events can be useful to understand binary evolution of massive stars in the Milky Way, while high-mass BH lenses can be useful to probe single stellar evolution.

    « less
  5. Abstract

    Long-duration gamma-ray bursts (lGRBs) originate in relativistic collimated outflows—jets—that drill their way out of collapsing massive stars. Accurately modeling this process requires realistic stellar profiles for the jets to propagate through and break out of. Most previous studies have used simple power laws or pre-collapse models for massive stars. However, the relevant stellar profile for lGRB models is in fact that of a star after its core has collapsed to form a compact object. To self-consistently compute such a stellar profile, we use the open-source code GR1D to simulate the core-collapse process for a suite of low-metallicity rotating massive stellar progenitors that have undergone chemically homogeneous evolution. Our models span a range of zero-age main-sequence (ZAMS) masses:MZAMS= 13, 18, 21, 25, 35, 40, and 45M. All of these models, at the onset of core-collapse, feature steep density profiles,ρrα, withα≈ 2.5, which would result in jets that are inconsistent with lGRB observables. We follow the collapses of four of the seven models until they form black holes (BHs) and the other three models until they form proto-neutron stars (PNSs). We find, across all models, that the density profile outside the newly formed BH or PNS is well represented by amore »flatter power law withα≈ 1.35–1.55. Such flat density profiles are conducive to the successful formation and breakout of BH-powered jets and are, in fact, required to reproduce observable properties of lGRBs. Future models of lGRBs should be initialized with shallower post-collapse stellar profiles, like those presented here, instead of the much steeper pre-collapse profiles that are typically used.

    « less