Improving the Detection of Gravitational-Wave Signals in Real Time

TL;DR

This thesis advances real-time gravitational-wave detection by combining enhanced noise modelling, glitch subtraction, and low-latency search optimisations. It introduces an exponential noise model and PSD-variation within PyCBC Live, improving ranking statistics and enabling greater sensitivity in live CBC searches, while ArchEnemy provides a glitch-subtraction pathway for scattered-light artifacts. The work also critically evaluates PyCBC Live’s early warning capabilities, proposing modifications to coincidence timing and phase-time-amplitude histograms to better capture pre-merger signals, particularly for electromagnetic follow-up of BNS events. Complementing these efforts, the study refines the PyCBC Live SNR optimiser—exploring differential evolution and PSO—to deliver modest but consistent gains in network SNR and faster sky localisation, with consideration of latency constraints. Collectively, these contributions enhance the responsiveness and reliability of gravitational-wave detections, enabling more timely multi-messenger observations and deeper insights into CBC populations, while identifying practical limits and future directions for live-search pipelines.

Abstract

This thesis presents advancements in the detection of gravitational waves from compact binary coalescences, utilising the most sensitive observatories constructed to date. The research focuses on enhancing gravitational-wave signal searches through the development of new tools and the application of existing methodologies to increase the sensitivity of live gravitational-wave searches. We introduced a novel noise artefact model, which enabled the identification and removal of glitches, thereby facilitating the recovery of previously missed gravitational-wave injections. This pioneering approach established a glitch search pipeline that adapted techniques typically used in gravitational-wave searches to address the unique characteristics of glitches. Additionally, we implemented an exponential noise model within the PyCBC Live search framework, significantly improving the detection ranking statistics for gravitational-wave signals and demonstrating the potential for substantial increases in detection sensitivity. Furthermore, we analysed and proposed enhancements for the PyCBC Live Early Warning search to maximise the detection of gravitational-wave events in the early warning regime. Our findings highlighted deficiencies in the current ranking statistic and led to recommendations for optimising coincidence timing windows and refining phase-time-amplitude histograms. These adjustments aim to increase the detection of gravitational-wave signals, particularly binary neutron star events, in early warning scenarios. The results underscore the importance of advancing search techniques in gravitational-wave astronomy, which can operate independently of detector improvements. By refining search methodologies, we enhance the capacity to detect a greater number of events, contributing significantly to our understanding of the Universe.

Figures (65)

• Figure 1: The effect of the two polarisations on a ring of test particles gw_polarisation_plots.
• Figure 2: An example Michelson laser interferometer detailing the main components of the advanced LIGO gravitational-wave observatory aLIGO:2015. Half the laser is reflected to the north arm cavity, physically defined by the input test mass (ITM$_{\text{N}}$) and the end test mass (ETM$_{\text{N}}$), and the other half is transmitted through to the east arm. Laser power builds up in the arm cavities through reflection between test masses before returning to the beam splitter, where there will be a total destructive interference of light when no gravitational-wave signal is present and therefore no signal will be measured at the photo-detector output. The power recycling mirror reflects laser light back into the detector arms to build up more power. The signal recycling mirror reflects light back into the arms with frequencies like those expected from gravitational-wave signals which resonates inside the cavity and increase the signal sensitive frequency power. Taken from IFO_diagram:2008.
• Figure 3: The angles of rotation to describe a gravitational wave from the source from (unprimed coordinates) to the detector frame (single primed coordinates) through the radiation frame (double primed coordinates). Taken from Brown_Thesis:2004.
• Figure 4: The evolution of a compact binary merger is illustrated through three distinct phases. The inspiral phase involves the two compact objects orbiting each other, with the orbit decaying as they emit gravitational waves. This phase is modelled using post-Newtonian theory. The merger phase occurs when the orbital radius decreases sufficiently for the objects to merge into a new compact object, modelled using numerical relativity. The ringdown phase represents the newly formed object's vibration as it settles into its final shape, modelled using perturbation theory. This image is adapted from IMR_plot:2016.
• Figure 5: Gating a very loud noise transient. The detector strain has been rescaled by a factor of $10^{21}$ and the glitch has a peak magnitude over $5$,$000$. The blue line shows the data before applying the Tukey window and the red shows the data after applying the Tukey window. Note the smooth decrease in data amplitude at the edges of the windowing function. Image taken from PyCBC:2016.