Utilizing anticoincidence veto in a search for gravitational-wave transients

TL;DR

The paper addresses noise glitches in gravitational-wave transient searches by exploiting temporal anticoincidence across geographically separated detectors. It develops an anticoincidence veto integrated into a PSO-based matched-filter BBH search, combining chi-squared and sine-Gaussian vetoes with a time-coincidence test to suppress non-astrophysical triggers. The results show backgrounds becoming Gaussian-like and injections being recovered more efficiently, especially for moderate-SNR signals, with minimal computational overhead. This approach enhances sensitivity for short-duration transients and is applicable to any coincident GW search across detector networks.

Abstract

We devise a technique to suppress the effect of noise transients occurring at gravitational-wave detectors based on temporal anticoincidence. Searches for gravitational-wave signals in the detector data are prone to spurious disturbances of terrestrial origin. The technique presented here benefits from the fact that such effects are generally non-coincident in time at geographically separated detectors. Therefore, abnormally loud triggers that are not time-coincident can be vetoed. We implement the veto technique in a matched-filter search for transient signals from binary black holes and observe search backgrounds to be generally close to the Gaussian limit. An improvement in the sensitivity of the search is demonstrated using simulated signals. The technique is expected to especially improve the detection efficiency of the search toward short duration transient signals.

Figures (9)

• Figure 1: Data blocks obtained as coincident observation times between LIGO Hanford (H1) and LIGO Livingston (L1) during their third Observing Run (O3). Here a total of about 30 days of observations is shown which resulted into approximately 14.5 days of coincident observing time.
• Figure 2: Triggers generated from the optimized templates obtained for the coincident data stretches. The SNR and the reweighted SNR for each trigger is plotted against its GPS timestamp. The triggers above the horizontal dashed line in grey (threshold of $\sim$ 4.0 on the SNR) are retained for the next stage of the analysis. Also, note that a significant number of triggers can already be discarded by comparing their SNR and reweighted SNR values.
• Figure 3: Impact of the anticoincidence veto on the distribution of detector triggers ranked by their reweighted SNRs. These exclude those that are caused due to the injections. We observe that the veto step discards a significant number of abnormally loud triggers from each of the detectors, resulting in a distribution which is close to one obtained from Gaussian noise. However, there are very few reasonably loud triggers failing in the anticoincidence test, which could possibly be of astrophysical origin.
• Figure 4: Self-consistency of the FAR estimates with and without the anticoincidence veto in Gaussian noise (left) and in instrument noise (right) from O3a. The effect of the veto is demonstrated on the same corresponding datasets split into smaller chunks using identical analysis configuration.
• Figure 5: Distribution of background events over ranking statistic with and without the anticoincidence veto applied. We observe that while in Gaussian noise (left), the veto has negligible effect on the distributions, the same has a large effect in generally obtaining Gaussian-like distributions in instrument noise (right).