Detection methods for non-Gaussian gravitational wave stochastic backgrounds

TL;DR

This paper develops an optimal detection method for non-Gaussian, intermittent gravitational wave backgrounds, showing it outperforms standard methods under certain conditions, with potential improvements in detection sensitivity.

Contribution

It introduces a maximum likelihood detection statistic tailored for non-Gaussian gravitational wave backgrounds, improving detection sensitivity over traditional cross-correlation methods.

Findings

Maximum likelihood statistic outperforms cross-correlation for non-Gaussian backgrounds.

Detection sensitivity improves by a factor of 1 to 3 in energy density.

Analysis assumes white Gaussian noise and collocated detectors, requiring further generalization.

Abstract

We address the issue of finding an optimal detection method for a discontinuous or intermittent gravitational wave stochastic background. Such a signal might sound something like popcorn popping. We derive an appropriate version of the maximum likelihood detection statistic, and compare its performance to that of the standard cross-correlation statistic both analytically and with Monte Carlo simulations. The maximum likelihood statistic performs better than the cross-correlation statistic when the background is sufficiently non-Gaussian. For both ground and space based detectors, this results in a gain factor, ranging roughly from 1 to 3, in the minimum gravitational-wave energy density necessary for detection, depending on the duty cycle of the background. Our analysis is exploratory, as we assume that the time structure of the events cannot be resolved, and we assume white, Gaussian noise in two collocated, aligned detectors. Before this detection method can be used in practice with real detector data, further work is required to generalize our analysis to accommodate separated, misaligned detectors with realistic, colored, non-Gaussian noise.