[Paper Review] Searching for gravitational wave echoes in GWTC-1 and O3 events

Yutong Wang, Yun-Song Piao|arXiv (Cornell University)|Oct 15, 2020

Pulsars and Gravitational Waves Research4 citations

TL;DR

This study uses Bayesian model selection and Markov Chain Monte Carlo (MCMC) analysis to search for gravitational wave echoes in all GWTC-1 and O3 gravitational wave events, using a phenomenological IMRE (Inspiral-Merger-Ringdown-Echo) waveform model. It finds no statistically significant evidence for echoes across the full event sets, with combined Bayes factors indicating less than 1σ preference for echo models, though a few individual events show weak hints at 1–2σ significance.

ABSTRACT

Gravitational wave (GW) echoes, if they exist, would be a probe to the near-horizon physics of black hole. In this brief report, we performed the Monte Carlo Markov Chain analysis to search for echo signal in all GWTC-1 and O3 GW events. We focus on the Insprial-Merger-Ringdown-Echo (IMRE) waveform, and apply the Bayesian model selection to compare the IMRE result with IMR's (no echo). We find no statistically significant ($<1σ$ combined) evidence for the GW echoes and only individual GW events with the echoes at $1\sim 2σ$ significance.

Motivation & Objective

• To investigate the presence of gravitational wave echoes in all GWTC-1 and O3 gravitational wave events as a probe of near-horizon physics beyond general relativity.
• To test whether the IMRE (Inspiral-Merger-Ringdown-Echo) waveform model provides a better fit than the standard IMR (no echo) model using Bayesian model selection.
• To assess the statistical significance of echo signals across multiple events, accounting for potential non-uniform echo intervals.
• To evaluate the sensitivity of current detectors to echo signals and identify promising candidates for future detection.

Proposed method

• The study employs a phenomenological IMRE waveform model that adds echo components to the standard IMR waveform, with echo amplitudes decaying as 1/(3+n) and intervals increasing linearly with a parameter r.
• The MCMC algorithm, implemented via the emcee3.0.2 package and MGWB code, explores the full parameter space including IMR parameters, echo amplitude ratio 𝒜, echo duration β, echo interval increment r, echo start time t_echo, and echo period Δt_echo.
• Bayesian model selection is applied using the log Bayes factor ln B₀₁ = ln[p(d|H₁,I)/p(d|H₀,I)] to compare the IMRE model (H₁) against the standard IMR model (H₀) for each event.
• The combined Bayes factor across all events is computed as the sum of individual log Bayes factors, allowing assessment of overall model preference.
• The analysis considers both constant and increasing echo interval models, with r ≠ 0 indicating potential signatures of exotic compact objects like wormholes.
• Posterior distributions for key parameters are visualized, e.g., for GW190425, showing 68% and 95% credible regions.

Experimental results

Research questions

• RQ1 Is there statistically significant evidence for gravitational wave echoes in the GWTC-1 and O3 gravitational wave event sets using Bayesian model comparison?
• RQ2 Do any individual GW events show a preference for the IMRE model over the standard IMR model, even if the combined sample does not?
• RQ3 How does the inclusion of non-uniform echo intervals (via parameter r) affect the detection significance of echo signals?
• RQ4 What is the overall preference for the echo model across all events, and how does it compare to the sensitivity of current detectors?
• RQ5 Can the observed weak hints in certain events (e.g., GW170817, GW190412) be confirmed with future, more sensitive detectors?

Key findings

• The combined Bayes factor for all GWTC-1 events is -1.49, corresponding to a 0.14σ preference for the IMR model, indicating no significant evidence for echoes.
• The combined Bayes factor for all O3 events is -1.08, corresponding to a 0.26σ preference for the IMR model, suggesting slightly better sensitivity in O3 compared to GWTC-1.
• GW151012 shows the strongest individual evidence with a log Bayes factor of 1.96, corresponding to 1.84σ significance for the IMRE model.
• GW170817 and GW170823 both show log Bayes factors of 0.55, corresponding to 1.10σ significance, indicating weak positive hints for echoes.
• No event shows evidence exceeding 2σ significance, and no event set shows a combined preference for the echo model at more than 1σ level.
• The results suggest that while current data do not support the existence of gravitational wave echoes, future detectors like aLIGO O5 or the Einstein Telescope may detect such signals if they exist.