Search for: All records

GWSkyNet-Multiis 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 byGWSkyNet-Multiin 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.

The two interferometric LIGO gravitational-wave observatories provide the most sensitive data to date to study the gravitational-wave universe. As part of a global network, they have completed their third observing run in which they observed many tens of signals from merging compact binary systems. It has long been known that a limiting factor in identifying transient gravitational-wave signals is the presence of transient non-Gaussian noise, which reduce the ability of astrophysical searches to detect signals confidently. Significant efforts are taken to identify and mitigate this noise at the source, but its presence persists, leading to the need for software solutions. Taking a set of transient noise artefacts categorised by the GravitySpy software during the O3a observing era, we produce parameterised population models of the noise projected into the space of astrophysical model parameters of merging binary systems. We compare the inferred population properties of transient noise artefacts with observed astrophysical systems from the GWTC2.1 catalogue. We find that while the population of astrophysical systems tend to have near equal masses and moderate spins, transient noise artefacts are typically characterised by extreme mass ratios and large spins. This work provides a new method to calculate the consistency of an observed candidate with a given class of noise artefacts. This approach could be used in assessing the consistency of candidates found by astrophysical searches (i.e. determining if they are consistent with a known glitch class). Furthermore, the approach could be incorporated into astrophysical searches directly, potentially improving the reach of the detectors, though only a detailed study would verify this.

Orbital eccentricity is a key signature of dynamical binary black hole formation. The gravitational waves from a coalescing binary contain information about its orbital eccentricity, which may be measured if the binary retains sufficient eccentricity near merger. Dedicated waveforms are required to measure eccentricity. Several models have been put forward, and show good agreement with numerical relativity at the level of a few percent or better. However, there are multiple ways to define eccentricity for inspiralling systems, and different models internally use different definitions of eccentricity, making it difficult to compare eccentricity measurements directly. In this work, we systematically compare two eccentric waveform models,SEOBNREandTEOBResumS, by developing a framework to translate between different definitions of eccentricity. This mapping is constructed by minimizing the relative mismatch between the two models over eccentricity and reference frequency, before evolving the eccentricity of one model to the same reference frequency as the other model. We show that for a given value of eccentricity passed toSEOBNRE, one must input a 20%–50% smaller value of eccentricity toTEOBResumSin order to obtain a waveform with the same empirical eccentricity. We verify this mapping by repeating our analysis for eccentric numerical relativity simulations, demonstrating thatTEOBResumSreports a correspondingly smaller value of eccentricity thanSEOBNRE.