Constraining Gravitational Wave Memory with Hierarchical Inference

Abstract

With the multitude of gravitational wave observations that have been made in the past ten years, probing the dynamical and nonlinear nature of strong gravity is becoming more and more feasible. One promising way to test the nonlinear nature of Einstein's theory of general relativity (GR) is through the gravitational wave null memory effect: a nonlinear prediction of GR which corresponds to initially comoving observers being permanently displaced due to a burst of gravitational radiation. Previous studies have shown that, while it is unlikely that the memory effect will be observed in a single event by the LIGO-Virgo-KAGRA (LVK) detectors, evidence for memory in the population of LVK events should be attainable after $\sim$2,000 gravitational wave detections. These works, however, largely relied on Bayes factors to perform their memory analyses: an approach that can depend sensitively on the analysis priors and, when naively multiplied across many events, can even favor incorrect conclusions. In this work, using the GWTC-5.0 catalog of binary black hole observations, we instead perform hierarchical Bayesian inference -- which is not subject to the issues associated with Bayes factors -- to measure the evidence for memory in current LVK observations. We find that we can constrain what we call the memory enhancement factor -- the constant appearing in front of the contribution to the strain from the supermomentum flux -- to $0.26_{-4.08}^{+4.09}$ (with $\pm$ values denoting the 68\% credible interval), consistent with its GR value of 1. We also forecast that $\sim$2,000 detections will be needed to constrain the memory enhancement factor away from zero at the $1\sigma$ level.