Electromagnetic Follow-up Research Articles

A rapid multi-modal parameter estimation technique for LISA

Abstract The Laser Interferometer Space Antenna (LISA) will observe gravitational-wave signals from a wide range of sources, including massive black hole binaries. Although numerous techniques have been developed to perform Bayesian inference for LISA, they are often computationally expensive; analyses often take at least $\sim 1$ month on a single CPU, even when using accelerated techniques. Not only does this make it difficult to concurrently analyse more than one gravitational-wave signal, it also makes it challenging to rapidly produce parameter estimates for possible electromagnetic follow-up campaigns. simple-pe was recently developed to produce rapid parameter estimates for gravitational-wave signals observed with ground-based gravitational-wave detectors. In this work, we extend simple-pe to produce rapid parameter estimates for LISA sources, including the effects of higher order multipole moments. We show that simple-pe infers the source properties of massive black hole binaries in zero-noise at least $\sim 100\times$ faster than existing techniques; $\sim 12$ hours on a single CPU. We further demonstrate that simple-pe can be applied before existing Bayesian techniques to mitigate biases in multi-modal parameter estimation analyses of MBHBs.

GLADEnet: A progressive web app for multi-messenger cosmology and electromagnetic follow-ups of gravitational-wave sources

Multi-messenger astronomy is an emerging field of research aimed at unravelling the physics governing astrophysical transients. GW170817 stands out as the first multi-messenger observation of the coalescence of a binary system of neutron stars, detected by the LIGO and Virgo gravitational-wave interferometers, along with space- and ground-based electromagnetic telescopes. It is a striking example of how multi-messenger observations significantly enhance our understanding of the physics of compact objects, relativistic outflows, and nucleosynthesis. It shows a new way of making cosmology and has the potential to resolve the tension between different measurements of the expansion rate of the Universe. To optimise multi-messenger observational strategies, to evaluate the efficiency of the searches for counterparts, and to identify the host galaxy of the source in a large sky localisation, information about the volumes of galaxies within the gravitational-wave localisation is of paramount importance. This requires the use of galaxy catalogues and appropriate knowledge of their completeness. Here, we describe a new interactive web tool namedGLADEnetthat allows us to identify catalogued galaxies and to assess the incompleteness of the catalogue of galaxies in real time across the gravitational-wave sky localisation. This measure is of particular importance when using catalogues such as the GLADE catalogue (Galaxy List for the Advanced Detector Era), which includes a collection of various catalogues that make completeness differ across different regions of the sky. We discuss the analysis steps to defining a completeness coefficient and provide a comprehensive guide on how to use the web app, detailing its functionalities. The app is geared towards managing the vast collection of over 22 million objects in GLADE. The completeness coefficient and the GLADE galaxy list will be disseminated in real time viaGLADEnet, powered by the Virtual Observatory (VO) standard and tools.

Explaining the GWSkyNet-Multi Machine Learning Classifier Predictions for Gravitational-wave Events

GWSkyNet-Multi is a machine learning model developed for the classification of candidate gravitational-wave events detected by the LIGO and Virgo observatories. The model uses limited information released in the low-latency Open Public Alerts to produce prediction scores indicating whether an event is a merger of two black holes (BHs), a merger involving a neutron star (NS), or a non-astrophysical glitch. This facilitates time-sensitive decisions about whether to perform electromagnetic follow-up of candidate events during LIGO-Virgo-KAGRA (LVK) observing runs. However, it is not well understood how the model is leveraging the limited information available to make its predictions. As a deep learning neural network, the inner workings of the model can be difficult to interpret, impacting our trust in its validity and robustness. We tackle this issue by systematically perturbing the model and its inputs to explain what underlying features and correlations it has learned for distinguishing the sources. We show that the localization area of the 2D sky maps and the computed coherence versus incoherence Bayes factors are used as strong predictors for distinguishing between real events and glitches. The estimated distance to the source is further used to discriminate between binary BH mergers and mergers involving NSs. We leverage these findings to show that events misclassified by GWSkyNet-Multi in LVK’s third observing run have distinct sky areas, coherence factors, and distance values that influence the predictions and explain these misclassifications. The results help identify the model’s limitations and inform potential avenues for further optimization.

Prospects for realtime characterization of core-collapse supernova and neutrino properties

Core-collapse supernovae (CCSNe) offer extremely valuable insights into the dynamics of galaxies. Neutrino time profiles from CCSNe, in particular, could reveal unique details about collapsing stars and particle behavior in dense environments. However, CCSNe in our galaxy and the Large Magellanic Cloud are rare and only one supernova neutrino observation has been made so far. To maximize the information obtained from the next Galactic CCSN, it is essential to combine analyses from multiple neutrino experiments in real time and transmit any relevant information to electromagnetic facilities within minutes. Locating the CCSN, in particular, is challenging, requiring disentangling CCSN localization information from observational features associated with the properties of the supernova progenitor and the physics of the neutrinos. Yet, being able to estimate the progenitor distance from the neutrino signal would be of great help for the optimisation of the electromagnetic follow-up campaign that will start soon after the propagation of the neutrino alert. Existing CCSN distance measurement algorithms based on neutrino observations hence rely on the assumption that neutrino properties can be described by the Standard Model. This paper presents a swift and robust approach to extract CCSN and neutrino physics information, leveraging diverse next-generation neutrino detectors to counteract potential measurement biases from Beyond the Standard Model effects.

Designing an Optimal Kilonova Search Using DECam for Gravitational-wave Events

We address the problem of optimally identifying all kilonovae detected via gravitational-wave emission in the upcoming LIGO/Virgo/KAGRA observing run, O4, which is expected to be sensitive to a factor of ∼7 more binary neutron star (BNS) alerts than previously. Electromagnetic follow-up of all but the brightest of these new events will require >1 m telescopes, for which limited time is available. We present an optimized observing strategy for the DECam during O4. We base our study on simulations of gravitational-wave events expected for O4 and wide-prior kilonova simulations. We derive the detectabilities of events for realistic observing conditions. We optimize our strategy for confirming a kilonova while minimizing telescope time. For a wide range of kilonova parameters, corresponding to a fainter kilonova compared to GW170817/AT 2017gfo, we find that, with this optimal strategy, the discovery probability for electromagnetic counterparts with the DECam is ∼80% at the nominal BNS gravitational-wave detection limit for O4 (190 Mpc), which corresponds to an ∼30% improvement compared to the strategy adopted during the previous observing run. For more distant events (∼330 Mpc), we reach an ∼60% probability of detection, a factor of ∼2 increase. For a brighter kilonova model dominated by the blue component that reproduces the observations of GW170817/AT 2017gfo, we find that we can reach ∼90% probability of detection out to 330 Mpc, representing an increase of ∼20%, while also reducing the total telescope time required to follow up events by ∼20%.

Updated Observing Scenarios and Multimessenger Implications for the International Gravitational-wave Networks O4 and O5

An advanced LIGO and Virgo’s third observing run brought another binary neutron star merger (BNS) and the first neutron-star black hole mergers. While no confirmed kilonovae were identified in conjunction with any of these events, continued improvements of analyses surrounding GW170817 allow us to project constraints on the Hubble Constant (H 0), the Galactic enrichment from r-process nucleosynthesis, and ultra-dense matter possible from forthcoming events. Here, we describe the expected constraints based on the latest expected event rates from the international gravitational-wave network and analyses of GW170817. We show the expected detection rate of gravitational waves and their counterparts, as well as how sensitive potential constraints are to the observed numbers of counterparts. We intend this analysis as support for the community when creating scientifically driven electromagnetic follow-up proposals. During the next observing run O4, we predict an annual detection rate of electromagnetic counterparts from BNS of () for the Zwicky Transient Facility (Rubin Observatory).

Optimization of the SVOM satellite strategy for the rapid follow-up of gravitational wave events

ABSTRACT The SVOM satellite, to be launched in early 2024, is primarily devoted to the multiwavelength observation of gamma-ray bursts and other higher-energy transients. Thanks to its onboard Microchannel X-ray Telescope and Visible-band Telescope, it is also very well adapted to the electromagnetic follow-up of gravitational wave events. We discuss the SVOM rapid follow-up strategy for gravitational wave trigger candidates provided by LIGO-Virgo-KAGRA. In particular, we make use of recent developments of galaxy catalogues adapted to the horizon of gravitational wave detectors to optimize the chance of counterpart discovery. We also take into account constraints specific to the SVOM platform. Finally, we implement the production of the SVOM observation plan following a gravitational wave alert and quantify the efficiency of several optimizations introduced in this work.

Panning for gold, but finding helium: Discovery of the ultra-stripped supernova SN 2019wxt from gravitational-wave follow-up observations

We present the results from multi-wavelength observations of a transient discovered during an intensive follow-up campaign of S191213g, a gravitational wave (GW) event reported by the LIGO-Virgo Collaboration as a possible binary neutron star merger in a low latency search. This search yielded SN 2019wxt, a young transient in a galaxy whose sky position (in the 80% GW contour) and distance (∼150 Mpc) were plausibly compatible with the localisation uncertainty of the GW event. Initially, the transient’s tightly constrained age, its relatively faint peak magnitude (Mi ∼ −16.7 mag), and the r-band decline rate of ∼1 mag per 5 days appeared suggestive of a compact binary merger. However, SN 2019wxt spectroscopically resembled a type Ib supernova, and analysis of the optical-near-infrared evolution rapidly led to the conclusion that while it could not be associated with S191213g, it nevertheless represented an extreme outcome of stellar evolution. By modelling the light curve, we estimated an ejecta mass of only ∼0.1 M⊙, with 56Ni comprising ∼20% of this. We were broadly able to reproduce its spectral evolution with a composition dominated by helium and oxygen, with trace amounts of calcium. We considered various progenitor channels that could give rise to the observed properties of SN 2019wxt and concluded that an ultra-stripped origin in a binary system is the most likely explanation. Disentangling genuine electromagnetic counterparts to GW events from transients such as SN 2019wxt soon after discovery is challenging: in a bid to characterise this level of contamination, we estimated the rate of events with a volumetric rate density comparable to that of SN 2019wxt and found that around one such event per week can occur within the typical GW localisation area of O4 alerts out to a luminosity distance of 500 Mpc, beyond which it would become fainter than the typical depth of current electromagnetic follow-up campaigns.

Neutron star-black hole mergers in next generation gravitational-wave observatories

Observations by the current generation of gravitational-wave detectors have been pivotal in expanding our understanding of the universe. Although tens of exciting compact binary mergers have been observed, neutron star-black hole (NSBH) mergers remained elusive until they were first confidently detected in 2020. The number of NSBH detections is expected to increase with sensitivity improvements of the current detectors and the proposed construction of new observatories over the next decade. In this work, we explore the NSBH detection and measurement capabilities of these upgraded detectors and new observatories using the following metrics: network detection efficiency and detection rate as a function of redshift, distributions of the signal-to-noise ratios, the measurement accuracy of intrinsic and extrinsic parameters, the accuracy of sky position measurement, and the number of early-warning alerts that can be sent to facilitate the electromagnetic follow-up. Additionally, we evaluate the prospects of performing multimessenger observations of NSBH systems by reporting the number of expected kilonova detections with the Vera C. Rubin Observatory and the Nancy Grace Roman Space Telescope. We find that as many as $\mathcal{O}(10)$ kilonovae can be detected by these two telescopes every year, depending on the population of the NSBH systems and the equation of state of neutron stars.

Dark photon superradiance: Electrodynamics and multimessenger signals

We study the electrodynamics of a kinetically mixed dark photon cloud that forms through superradiance around a spinning black hole, and design strategies to search for the resulting multimessenger signals. A dark photon superradiance cloud sources a rotating dark electromagnetic field which, through kinetic mixing, induces a rotating visible electromagnetic field. Standard model charged particles entering this field initiate a transient phase of particle production that populates a plasma inside the cloud and leads to a system which shares qualitative features with a pulsar magnetosphere. We study the electrodynamics of the dark photon cloud with resistive magnetohydrodynamics methods applicable to highly magnetized plasma, adapting techniques from simulations of pulsar magnetospheres. We identify turbulent magnetic field reconnection as the main source of dissipation and electromagnetic emission, and compute the peak luminosity from clouds around solar-mass black holes to be as large as $10^{43}$ erg/s for open dark photon parameter space. The emission is expected to have a significant X-ray component and is potentially periodic, with period set by the dark photon mass. The luminosity is comparable to the brightest X-ray sources in the Universe, allowing for searches at distances of up to hundreds of Mpc with existing telescopes. We discuss observational strategies, including targeted electromagnetic follow-ups of solar-mass black hole mergers and targeted continuous gravitational wave searches of anomalous pulsars.

Neutrino follow-up with the Zwicky transient facility: results from the first 24 campaigns

ABSTRACT The Zwicky Transient Facility (ZTF) performs a systematic neutrino follow-up programme, searching for optical counterparts to high-energy neutrinos with dedicated Target-of-Opportunity (ToO) observations. Since first light in March 2018, ZTF has taken prompt observations for 24 high-quality neutrino alerts from the IceCube Neutrino Observatory, with a median latency of 12.2 h from initial neutrino detection. From two of these campaigns, we have already reported tidal disruption event (TDE) AT 2019dsg and likely TDE AT 2019fdr as probable counterparts, suggesting that TDEs contribute >7.8 per cent of the astrophysical neutrino flux. We here present the full results of our programme through to December 2021. No additional candidate neutrino sources were identified by our programme, allowing us to place the first constraints on the underlying optical luminosity function of astrophysical neutrino sources. Transients with optical absolutes magnitudes brighter that −21 can contribute no more than 87 per cent of the total, while transients brighter than −22 can contribute no more than 58 per cent of the total, neglecting the effect of extinction and assuming they follow the star formation rate. These are the first observational constraints on the neutrino emission of bright populations such as superluminous supernovae. None of the neutrinos were coincident with bright optical AGN flares comparable to that observed for TXS 0506+056/IC170922A, with such optical blazar flares producing no more than 26 per cent of the total neutrino flux. We highlight the outlook for electromagnetic neutrino follow-up programmes, including the expected potential for the Rubin Observatory.

Deep learning detection and classification of gravitational waves from neutron star-black hole mergers

The Laser Interferometer Gravitational-Wave Observatory (LIGO) and Virgo Interferometer Collaborations have now detected all three classes of compact binary mergers: binary black hole (BBH), binary neutron star (BNS), and neutron star-black hole (NSBH). For coalescences involving neutron stars, the simultaneous observation of gravitational and electromagnetic radiation produced by an event, has broader potential to enhance our understanding of these events, and also to probe the equation of state (EOS) of dense matter. However, electromagnetic follow-up to gravitational wave (GW) events requires rapid real-time detection and classification of GW signals, and conventional detection approaches are computationally prohibitive for the anticipated rate of detection of next-generation GW detectors. In this work, we present the first deep learning based results of classification of GW signals from NSBH mergers in real LIGO data. We show for the first time that a deep neural network can successfully distinguish all three classes of compact binary mergers and separate them from detector noise. Specifically, we train a convolutional neural network (CNN) on ∼500,000 data samples of real LIGO noise with injected BBH, BNS, and NSBH GW signals, and we show that our network has high sensitivity and accuracy. Most importantly, we successfully recover the two confirmed NSBH events to-date (GW191219 and GW200115) and the two confirmed BNS mergers to-date (GW170817 and GW190425), together with ∼90% of all BBH candidate events from the third Gravitational Wave Transient Catalog, GWTC-3. These results are an important step towards low-latency real-time GW detection, enabling multi-messenger astronomy.

Prospects for detecting and localizing short-duration transient gravitational waves from glitching neutron stars without electromagnetic counterparts

Neutron stars are known to show accelerated spin-up of their rotational frequency called a glitch. Highly magnetized rotating neutron stars (pulsars) are frequently observed by radio telescopes (and in other frequencies), where the glitch is observed as irregular arrival times of pulses which are otherwise very regular. A glitch in an isolated neutron star can excite the fundamental (f)-mode oscillations which can lead to gravitational wave generation. Electromagnetic observations of pulsars (and hence pulsar glitches) require the pulsar to be oriented so that the jet is pointed toward the detector, but this is not a requirement for gravitational wave emission which is more isotropic and not jetlike. Hence, gravitational wave observations have the potential to uncover nearby neutron stars where the jet is not pointed towards the Earth. In this work, we study the prospects of finding glitching neutron stars using a generic all-sky search for short-duration gravitational wave transients. The analysis covers the high-frequency range from $1-4$ kHz of LIGO-Virgo detectors for signals up to a few seconds. We set upper limits for the third observing run of the LIGO-Virgo detectors and present the prospects for upcoming observing runs of LIGO, Virgo, KAGRA, and LIGO India. We find the detectable glitch size will be around $10^{-5}$ Hz for the fifth observing run for pulsars with spin frequencies and distances comparable to the Vela pulsar. We also present the prospects of localizing the direction in the sky of these sources with gravitational waves alone, which can facilitate electromagnetic follow-up. We find that for the five detector configuration, the localization capability for a glitch size of $10^{-5}$ Hz is around $132\,\mathrm{deg}^{2}$ at $1\sigma$ confidence for $50\%$ of events with distance and spin frequency as that of Vela.

Premerger localization of intermediate mass binary black holes with LISA and prospects of joint observations with Athena and LSST

The planned Laser Interferometric Space Antenna (LISA) will be able to detect gravitational waves (GWs) from intermediate mass binary black holes (IMBBHs) in the mass range $\sim 10^{2} \mbox{-} 10^{4}\,M_{\odot}$ up to a redshift $z\sim20$. Modulation effects due to orbital motion of LISA around the Sun facilitate precise premerger localization of the sources, which in turn would help in electromagnetic (EM) followups. In this work, we calculate the uncertainties in sky position, luminosity distance, and time of coalescence as a function of time to coalescence. For representative masses of the IMBBHs, we synthesize a population of binaries uniformly located and oriented on a sphere of radius 3 Gpc and compute the projected parameter measurement uncertainties using the Fisher information matrix. We find that for systems with a total mass of $10^3\,M_{\odot}$, the errors in the sky position and luminosity distance are $\sim 0.4\,\text{deg}^2$ and $\sim 6\%$, respectively, one day prior to coalescence. The coalescence time can be predicted with an uncertainty $\lesssim 10$ sec, one day before coalescence. We also find that for $10^3\,M_{\odot}$, around $40\%$ ($100\%$) of the population has a source localization that is smaller than the field of view of Athena (LSST) one day before the merger. These extremely precise measurements can be used to alert ground-based GW detectors and EM telescopes about the time and location of these mergers. We also discuss mechanisms that may produce EM emission from IMBBH mergers and study its detectability using the planned Legacy Survey of Space and Time (LSST) in the optical and Athena in the x-ray bands. Detection of an EM transient may provide us vital clues about the environments where these mergers occur and the distance estimation can pave the way for cosmography.

Investigating GW190425 with numerical-relativity simulations

The third observing run of the LIGO-Virgo Collaboration has resulted in many gravitational wave detections, including the binary neutron star merger GW190425. However, none of these events have been accompanied with an electromagnetic transient found during extensive follow-up searches. In this article, we perform new numerical-relativity simulations of binary neutron star and black hole-neutron star systems that have a chirp mass consistent with GW190425. Assuming that the GW190425's sky location was covered with sufficient accuracy during the electromagnetic follow-up searches, we investigate whether the nondetection of the kilonova is compatible with the source parameters estimated through the gravitational-wave analysis and how one can use this information to place constraints on the properties of the system. Our simulations suggest that GW190425 is incompatible with an unequal mass binary neutron star merger with a mass ratio $q<0.8$ when considering stiff or moderately stiff equations of state if the binary was face on and covered by the observation. Our analysis shows that a detailed observational result for kilonovae will be useful to constrain the mass ratio of binary neutron stars in future events.

The lure of sirens: joint distance and velocity measurements with third-generation detectors

ABSTRACT The next generation of detectors will detect gravitational waves from binary neutron stars at cosmological distances, for which around a 1000 electromagnetic follow-ups may be observed per year. So far, most work devoted to the expected cosmological impact of these standard sirens employed them only as distance indicators. Only recently their use as tracers of clustering, similar to what already proposed for supernovae, has been studied. Focusing on the expected specifications of the Einstein Telescope (ET), we forecast here the performance on cosmological parameters of future standard sirens as both distance and density indicators, with emphasis on the linear perturbation growth index and on spatial curvature. We improve upon previous studies in a number of ways: a more detailed analysis of available telescope time, the inclusion of more cosmological and nuisance parameters, the Alcock–Paczynski correction, the use of sirens also as both velocity and density tracers, and a more accurate estimation of the distance posterior. We find that the analysis of the clustering of sirens improves the constraints on H0 by 30 per cent and on Ωk0 by over an order of magnitude, with respect to their use merely as distance indicators. With 5 yr of joint ET and Rubin Observatory follow-ups we could reach precision of 0.1 km s−1 Mpc−1 in H0 and 0.02 in Ωk0 using only data in the range 0 < z < 0.5. We also find that the use of sirens as tracers of density, and not only velocity, yields good improvements on the growth of structure constraints.

Multi-messenger Observations of Binary Neutron Star Mergers in the O4 Run

We present realistic expectations for the number and properties of neutron star binary mergers to be detected as multi-messenger sources during the upcoming fourth observing run (O4) of the LIGO-Virgo-KAGRA gravitational-wave (GW) detectors, with the aim of providing guidance for the optimization of observing strategies. Our predictions are based on a population synthesis mode, which includes the GW signal-to-noise ratio, the kilonova (KN) optical and near-infrared light curves, the relativistic jet gamma-ray burst (GRB) prompt emission peak photon flux, and the afterglow light curves in radio, optical, and X-rays. Within our assumptions, the rate of GW events to be confidently detected during O4 is yr−1 (calendar year), 78% of which will produce a KN, and a lower 52% will also produce a relativistic jet. The typical depth of current optical electromagnetic search and follow-up strategies is still sufficient to detect most of the KNæ in O4, but only for the first night or two. The prospects for detecting relativistic jet emission are not promising. While closer events (within z ≲ 0.02) will likely still have a detectable cocoon shock breakout, most events will have their GRB emission (both prompt and afterglow) missed unless seen under a small viewing angle. This reduces the fraction of events with detectable jets to 2% (prompt emission, serendipitous) and 10% (afterglow, deep radio monitoring), corresponding to detection rates of and yr−1, respectively. When considering a GW subthreshold search triggered by a GRB detection, our predicted rate of joint GW+GRB prompt emission detections increases up to a more promising yr−1.

Foraging with MUSHROOMS: A Mixed-integer Linear Programming Scheduler for Multimessenger Target of Opportunity Searches with the Zwicky Transient Facility

Electromagnetic follow-up of gravitational-wave detections is very resource intensive, taking up hours of limited observation time on dozens of telescopes. Creating more efficient schedules for follow-up will lead to a commensurate increase in counterpart location efficiency without using more telescope time. Widely used in operations research and telescope scheduling, mixed-integer linear programming is a strong candidate to produce these higher-efficiency schedules, as it can make use of powerful commercial solvers that find globally optimal solutions to provided problems. We detail a new target-of-opportunity scheduling algorithm designed with Zwicky Transient Facility in mind that uses mixed-integer linear programming. We compare its performance to gwemopt, the tuned heuristic scheduler used by the Zwicky Transient Facility and other facilities during the third LIGO–Virgo gravitational-wave observing run. This new algorithm uses variable-length observing blocks to enforce cadence requirements and to ensure field observability, along with having a secondary optimization step to minimize slew time. We show that by employing a hybrid method utilizing both this scheduler and gwemopt, the previous scheduler used, in concert, we can achieve an average improvement in detection efficiency of 3%–11% over gwemopt alone for a simulated binary neutron star merger data set consistent with LIGO–Virgo’s third observing run, highlighting the potential of mixed-integer target of opportunity schedulers for future multimessenger follow-up surveys.

Electromagnetic follow-up observations of binary neutron star mergers with early warnings from decihertz gravitational-wave observatories

ABSTRACT We investigate the prospects of electromagnetic follow-up observations for binary neutron star (BNS) mergers, with the help of early warnings from decihertz gravitational-wave (GW) observatories, B-DECIGO and DO-Optimal. Extending the previous work, we not only give quick assessments of joint short γ-ray burst (sGRB) detection rates for different γ-ray satellites and BNS population models, but also elaborate on the analyses and results on multiband kilonova detections for survey telescopes with different limiting magnitudes. During an assumed 4-yr mission time for decihertz GW observatories, we find that for the goals of electromagnetic follow-ups, DO-Optimal performs better than B-DECIGO as a whole on the detection rate, and has a larger detectable distance for joint sGRB/kilonova searches. Taking the lognormal population model for BNS mergers and a 1-d early-warning time as an example, we discuss the accuracy in localization and timing, as well as the redshift distributions for various synergy observations with electromagnetic facilities and decihertz GW detectors. Based on our analyses, we propose a feasible ‘wait-for’ pattern as a novel detecting mode for future multimessenger astrophysics.

GWSkyNet-Multi: A Machine-learning Multiclass Classifier for LIGO–Virgo Public Alerts

Abstract Compact object mergers which produce both detectable gravitational waves and electromagnetic (EM) emission can provide valuable insights into the neutron star equation of state, the tension in the Hubble constant, and the origin of the r-process elements. However, EM follow-up of gravitational wave sources is complicated by false-positive detections, and the transient nature of the associated EM emission. GWSkyNet-Multi is a machine learning model that attempts facilitate EM follow-up by providing real-time predictions of the source of a gravitational wave detection. The model uses information from Open Public Alerts (OPAs) released by LIGO–Virgo within minutes of a gravitational wave detection. GWSkyNet was introduced in Cabero et al. as a binary classifier and uses the OPA skymaps to classify sources as either astrophysical or as glitches. In this paper, we introduce GWSkyNet-Multi, an extension of GWSkyNet which further distinguishes sources as binary black hole mergers, mergers involving a neutron star, or non-astrophysical glitches. GWSkyNet-Multi is a sequence of three one-versus-all classifiers trained using a class-balanced and physically motivated source mass distribution. Training on this data set, we obtain test set accuracies of 93.7% for binary black hole-versus-all, 94.4% for neutron star-versus-all, and 95.1% for glitch-versus-all. We obtain an overall accuracy of 93.4% using a hierarchical classification scheme. Furthermore, we correctly identify 36 of the 40 gravitational wave detections from the first half of LIGO–Virgo’s third observing run (O3a) and present predictions for O3b sources. As gravitational wave detections increase in number and frequency, GWSkyNet-Multi will be a powerful tool for prioritizing successful EM follow-up.