Virgo detector characterization and data quality: results from the O3 run

Abstract

The Advanced Virgo detector has contributed with its data to the rapid growth of the number of detected GW signals in the past few years, alongside the two Advanced LIGO instruments. First during the last month of the Observation Run 2 (O2) in August 2017 (with, most notably, the compact binary mergers GW170814 and GW170817), and then during the full Observation Run 3 (O3): an 11 months data taking period, between April 2019 and March 2020, that led to the addition of 79 events to the catalog of transient GW sources maintained by LIGO, Virgo and now KAGRA. These discoveries and the manifold exploitation of the detected waveforms benefit from an accurate characterization of the quality of the data, such as continuous study and monitoring of the detector noise sources. These activities, collectively named detector characterization and data quality or DetChar, span the whole workflow of the Virgo data, from the instrument front-end hardware to the final analyses. They are described in detail in the following article, with a focus on the results achieved by the Virgo DetChar group during the O3 run. Concurrently, a companion article describes the tools that have been used by the Virgo DetChar group to perform this work.

Neural network time-series classifiers for gravitational-wave searches in single-detector periods

“…The distributions obtained for the noise or glitch classes exhibit maxima at zero for the TCN and IT classifiers, while the maximum is shifted to around 0.1 for CNN. Moving from the peak to P s higher values, the distributions show a monotonic decay until the last bin P s = 1 where distributions exhibit an count increase which is more prominent for TCN 15 . The TCN classifier appears to reach the lowest background ≲10 −2 in normalised count units.…”

“…Since our objective is to achieve high-confidence classification, we are primarily interested in the immediate vicinity of P s = 1. This motivates us to reparameterise the P s statistic as 15 We relate this increase in the background count to an issue with the floating-point precision. λ := − log 10 (1 − P s ).…”

“…To make robust detection statements, those pipelines have to address a major difficulty: the presence in the data of short-duration noise artefacts, often called 'instrumental glitches' [14,15], that occur sporadically in the data and that can mimic the GW signal [16,17]. A very powerful tool to discriminate the signal from noise glitches is time coincidence across two or more separate detectors (see [18] for a discussion on multi-detector noise rejection techniques).…”

Waves in a forest: a random forest classifier to distinguish between gravitational waves and detector glitches

“…The priors are similar to ones used in Ashton et al [49], including standard priors used in GW astronomy, as defined in Veitch et al [47]. The non standard priors that we use are for chirp mass, for which we use a broad uniform prior in [8,200] M ⊙ to encompass all our simulated GWs under one common prior. For the luminosity distance, we use the standard uniform in source frame prior between [207 500]Mpc.…”

“…Glitches can exhibit a range of morphologies, and their origin is not necessarily known. In cases where the cause has been identified, the associated glitches can be eliminated [4,7,8]. Despite these efforts, a large number of glitches still persist in the detector data [1][2][3].…”