Aperture synthesis for gravitational-wave data analysis: Deterministic Sources

TL;DR

This paper develops a method for combining data from multiple gravitational-wave detectors to improve detection sensitivity for known waveform sources, accounting for correlated noise and outperforming traditional coincidence methods.

Contribution

It derives the likelihood function and optimal matched filter for joint detector data analysis, including correlated and non-Gaussian noise scenarios, advancing gravitational-wave data analysis techniques.

Findings

Optimal filtering outperforms coincidence analysis in detection efficiency.

Method accounts for correlated noise between detectors.

Approach is effective even with non-Gaussian noise.

Abstract

Gravitational wave detectors now under construction are sensitive to the phase of the incident gravitational waves. Correspondingly, the signals from the different detectors can be combined, in the analysis, to simulate a single detector of greater amplitude and directional sensitivity: in short, aperture synthesis. Here we consider the problem of aperture synthesis in the special case of a search for a source whose waveform is known in detail: \textit{e.g.,} compact binary inspiral. We derive the likelihood function for joint output of several detectors as a function of the parameters that describe the signal and find the optimal matched filter for the detection of the known signal. Our results allow for the presence of noise that is correlated between the several detectors. While their derivation is specialized to the case of Gaussian noise we show that the results obtained are, in fact, appropriate in a well-defined, information-theoretic sense even when the noise is non-Gaussian in character. The analysis described here stands in distinction to ``coincidence analyses'', wherein the data from each of several detectors is studied in isolation to produce a list of candidate events, which are then compared to search for coincidences that might indicate common origin in a gravitational wave signal. We compare these two analyses --- optimal filtering and coincidence --- in a series of numerical examples, showing that the optimal filtering analysis always yields a greater detection efficiency for given false alarm rate, even when the detector noise is strongly non-Gaussian.