Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Search for gravitational-wave bursts in the third Advanced LIGO-Virgo run with coherent WaveBurst enhanced by machine learning

Marek J. Szczepańczyk, Francesco Salemi, Sophie Bini, Tanmaya Mishra, Gabriele Vedovato, V. Gayathri, Imre Bartos, Shubhagata Bhaumik, Marco Drago, Odysse Halim, Claudia Lazzaro, Andrea Miani, Edoardo Milotti, Giovanni A. Prodi, Shubhanshu Tiwari, and Sergey Klimenko
Phys. Rev. D 107, 062002 – Published 6 March 2023
PDFHTMLExport Citation
Abstract
Authors
Article Text
  • INTRODUCTION
  • METHOD
  • SIGNAL MODELS
  • RESULTS
  • CONCLUSIONS
  • ACKNOWLEDGMENTS
  • APPENDICES
    • References

    Abstract

    This paper presents a search for generic short-duration gravitational-wave (GW) transients (or GW bursts) in the data from the third observing run of Advanced LIGO and Advanced Virgo. We use a coherent WaveBurst (cWB) pipeline enhanced with a decision-tree classification algorithm for more efficient separation of GW signals from noise transients. The machine-learning (ML) algorithm is trained on a representative set of noise events and a set of simulated stochastic signals that are not correlated with any known signal model. This training procedure preserves the model-independent nature of the search. We demonstrate that the ML-enhanced cWB pipeline can detect GW signals at a larger distance than previous model-independent searches. The sensitivity improvements are achieved across the broad spectrum of simulated signals, with the goal of testing the robustness of this model-agnostic search. At a false-alarm rate of one event per century, the detectable signal amplitudes are reduced up to almost an order of magnitude, most notably for the single-cycle signal morphologies. This ML-enhanced pipeline also improves the detection efficiency of compact binary mergers in a wide range of masses, from stellar mass to intermediate-mass black holes, both with circular and elliptical orbits. After excluding previously detected compact binaries, no new gravitational-wave signals are observed for the twofold Hanford-Livingston and the threefold Hanford-Livingston-Virgo detector networks. With the improved sensitivity of the all-sky search, we obtain the most stringent constraints on the isotropic emission of gravitational-wave energy from short-duration burst sources.

    • Figure
    • Figure
    • Figure
    • Figure
    • Figure
    • Figure
    • Figure
    • Received 15 November 2022
    • Accepted 1 February 2023

    DOI:https://doi.org/10.1103/PhysRevD.107.062002

    © 2023 American Physical Society

    Physics Subject Headings (PhySH)

    1. Research Areas
    Gravitational wave detectionGravitational waves
    Gravitation, Cosmology & Astrophysics

    Authors & Affiliations

    Marek J. Szczepańczyk1, Francesco Salemi2,3,*, Sophie Bini2,3, Tanmaya Mishra1, Gabriele Vedovato4, V. Gayathri1, Imre Bartos1, Shubhagata Bhaumik1, Marco Drago5,6, Odysse Halim7,8, Claudia Lazzaro4,9, Andrea Miani2,3, Edoardo Milotti7,8, Giovanni A. Prodi10,3, Shubhanshu Tiwari11, and Sergey Klimenko1

    • 1Department of Physics, University of Florida, PO Box 118440, Gainesville, Florida 32611-8440, USA
    • 2Università di Trento, Dipartimento di Fisica, I-38123 Povo, Trento, Italy
    • 3INFN, Trento Institute for Fundamental Physics and Applications, I-38123 Povo, Trento, Italy
    • 4Università di Padova, Dipartimento di Fisica e Astronomia, I-35131 Padova, Italy
    • 5Università di Roma La Sapienza, I-00185 Roma, Italy
    • 6INFN, Sezione di Roma, I-00185 Roma, Italy
    • 7Dipartimento di Fisica, Università di Trieste, I-34127 Trieste, Italy
    • 8INFN, Sezione di Trieste, I-34127 Trieste, Italy
    • 9INFN, Sezione di Padova, I-35131 Padova, Italy
    • 10Università di Trento, Dipartimento di Matematica, I-38123 Povo, Trento, Italy
    • 11Physik-Institut, University of Zurich, Winterthurerstrasse 190, 8057 Zurich, Switzerland

    • *francesco.salemi@unitn.it

    Article Text (Subscription Required)

    Click to Expand

    References (Subscription Required)

    Click to Expand
    Issue

    Vol. 107, Iss. 6 — 15 March 2023

    Reuse & Permissions
    Access Options
    CHORUS

    Article Available via CHORUS

    Download Accepted Manuscript
    Author publication services for translation and copyediting assistance advertisement

    Authorization Required

    Log In
    Other Options
    ×

    Images

    • Figure 1
      Figure 1

      Cumulative number of events versus iFAR found by the standard (brown dashed line) and ML-enhanced searches (blue dashed line) in O3. The red solid line shows the expected mean value of the background and the shaded regions are the 1σ, 2σ, and 3σ Poisson uncertainty intervals. Left: results for the HL network. At iFAR≥1  year, the ML-enhanced search detected 16 CBC events compared to 14 events for the standard search. Right: results for the HLV network, the standard search was not performed, while 10 events are detected with iFAR≥1  year. In both panels, the loudest events’ significance saturates due to the limited amount of background available for testing (1/2 was already used for training), i.e., roughly 500 (300) years for the HL (HLV) network. After removing the known CBC events (continuous lines), the ML-enhanced search reports a null result for both networks.

      Reuse & Permissions
    • Figure 2
      Figure 2

      Resulting hrss50 achieved with cWB with standard postproduction veto procedure (darker colors) and with ML-enhanced cWB (lighter colors) for the HL network on full O3 and at iFAR≥100  years. The waveforms reported are a subset of those listed in Table 1: ad-hoc signals ordered according to central frequency (red), core-collapse supernovae (green), ringdown waveforms (blue), and cosmic strings (yellow). The values on the top show the reduction factor on hrss50 with respect to the standard search; hrss50 ordinate scale decreases going upwards.

      Reuse & Permissions
    • Figure 3
      Figure 3

      Radiated energy in GWs at 50% detection efficiency and iFAR≥100  years for a source distance of 10 kpc. The ML-enhanced cWB improves the constraints across the frequency spectrum for all tested morphologies.

      Reuse & Permissions
    • Figure 4
      Figure 4

      Detection efficiency vs distance for sample supernova waveforms, for HL network at iFAR≥100  years. The ML-enhanced search improves the detection distance at 50% detection efficiency; the probability of detections at a closer distance increases significantly.

      Reuse & Permissions
    • Figure 5
      Figure 5

      Sensitivity volume obtained with cWB standard postproduction veto procedure (darker colors) and with ML-enhanced cWB (lighter colors) for HL network on full O3 data, at iFAR≥100  years. The ordinate reports relative sensitive volumes normalized by 4πr03, where r0 is the distance at which the source emits the reference amplitude hrss=10−22  Hz−1/2. We use this standard siren value across all reported signal models to highlight dependencies on signal morphology. The values on the top show the gain in the space volume VXGB/VSTD. From left to right, the waveforms reported are ad-hoc signals ordered according to frequency (red), ringdown waveforms (blue), and cosmic strings (yellow).

      Reuse & Permissions
    • Figure 6
      Figure 6

      Detection efficiency versus iFAR for BBH (black), IMBH (red), and eBBH (blue). ML-enhanced cWB (solid lines) shows an increase in detection efficiency with respect to the standard search (dashed lines).

      Reuse & Permissions
    • Figure 7
      Figure 7

      Sensitivity volume obtained with cWB standard postproduction veto procedure (darker colors) and with cWB ML-enhanced (lighter colors) for HLV network during full O3 run, at iFAR=100 years. The values above each column show the space volume gain VXGB/VSTD. From left to right, the waveforms reported are: ad-hoc signals ordered according to frequency (red), ringdown waveforms (blue), and cosmic strings (yellow).

      Reuse & Permissions
    ×

    Sign up to receive regular email alerts from Physical Review D

    Log In

    Cancel
    ×

    Search

    Article Lookup

    Paste a citation or DOI
    Enter a citation
    ×