Detection of the Permanent Strain Offset Component of Gravitational-Wave Memory in Black Hole Mergers

TL;DR

This paper introduces a new method for detecting the gravitational-wave memory effect from black hole mergers by directly measuring permanent space-time strain offsets, simplifying analysis and focusing on the most informative signal features.

Contribution

It presents a novel, simpler approach to detect gravitational-wave memory that avoids complex modeling of the full signal, applied to LIGO/Virgo data with partial success.

Findings

Possible detections at 2σ-4σ significance

Many upper limits established

Probability of false detection around 0.1

Abstract

We propose a novel approach to detecting the elusive gravitational-wave memory predicted by general relativity to accompany black hole mergers: direct measurement of the permanent space-time strain offset. Compared to previous techniques modeling and disentangling both the "chirp" and memory signals, this approach has several advantages: it targets the feature of the signal carrying nearly all its Shannon information, has great simplicity, circumvents the need for precise modeling of the time evolution of all components of the gravitational wave signal, and uses only data largely free of the more complicated chirp signal. The frequency spectrum of the predicted memory signal is roughly similar to that of the chirp signal. However its inclusion of lower frequencies, where noise and data calibration are problematic, makes detection difficult but not impossible. We applied this novel analysis, implemented with a template-like algorithm, to a selection of 67 observations of 41 black hole mergers in the LIGO/Virgo Gravitational Wave Transient Catalog. Statistical significance was assessed by analyzing many time-shifted intervals. The result: a few possible detections ($2\sigma-4\sigma$) and many upper limits. The probability that a random ensemble of 67 strain time series, with the same noise but no memory signals, will yield a particular figure-of-merit computed for the actual data is approximately 0.1. Several validation checks proved useless, partly due to large measurement and theoretical uncertainties, so these results should be viewed with reservation. Appendices contain MatLab code for various operations, including an algorithm for the complex Fourier transform of arbitrarily spaced data.