GW231123: Overlapping Gravitational Wave Signals?

This paper proposes that the gravitational wave event GW231123, previously interpreted as a single massive black hole merger, is more likely the result of two overlapping signals—potentially caused by gravitational lensing—which resolves significant discrepancies in source property measurements across different waveform models.

The Gist

The Big Mystery: A Heavyweight Champion with a Split Personality

Imagine the LIGO and Virgo detectors as incredibly sensitive microphones listening to the "sound" of the universe. Recently, they heard a very loud "chirp" from two black holes smashing together. This event, named GW231123, was special because the black holes were massive—so heavy that they broke the usual rules of how stars are supposed to die.

However, when scientists tried to measure the details of this crash (how heavy the black holes were, how fast they were spinning, and how far away they were), they hit a wall. It was like asking three different experts to describe the same car accident, and they all gave completely different stories. One said the car was red; another said it was blue. One said it was going 50 mph; another said 100 mph.

In the world of gravitational waves, these "experts" are called waveform models. They are complex computer programs that try to translate the raw sound data into physical facts. For GW231123, these models disagreed so much that the scientists couldn't trust the results. Something was wrong with the data, or the models were missing something.

The Detective's Theory: Two Songs Playing at Once

The authors of this paper proposed a new theory to solve the mystery: What if we weren't listening to just one crash, but two?

Imagine you are in a room where two people are singing different songs at the exact same time. If you try to figure out the lyrics of just one song without realizing the other is there, you will get confused. You might think the singer is singing a weird note, or that the song is longer than it really is.

The team tested this idea. They built a new "listening model" that assumed two separate gravitational wave signals were overlapping in the data, rather than just one.

The Result:
When they used this "two-songs" model, the confusion disappeared.

• The different computer models (the experts) finally agreed on the details of the main crash.
• The "two-songs" model fit the data much better than the "one-song" model. In statistical terms, the evidence was thousands of times stronger for the two-signal idea.

The Twist: Is it Two Crashes or a Cosmic Mirror?

So, does this mean two separate black hole collisions happened at the exact same time?

The authors ran the numbers on how likely this is. They calculated the odds of two massive black hole crashes happening within a fraction of a second of each other by pure chance. The result? It is extremely unlikely. It's like flipping a coin and getting heads a million times in a row. The universe just doesn't produce enough of these heavy crashes for them to randomly overlap like this.

However, the authors found a fascinating clue: The two "signals" they recovered looked almost identical. They had the same mass, the same spin, and they came from the same spot in the sky. They were separated by only a tiny fraction of a second (20 milliseconds).

This led to a new, more exciting possibility: Gravitational Lensing.

Think of a massive object (like a giant galaxy) sitting between the black holes and Earth. This object acts like a cosmic magnifying glass. It bends the light (or in this case, the gravitational waves) coming from the black holes. Sometimes, this bending creates two "images" of the same event. If the two images arrive at Earth almost at the same time, they look exactly like two overlapping signals.

The Verdict

The paper concludes that while it is highly unlikely that two different black hole collisions happened at the same time, the data strongly suggests that GW231123 might be a single event that was "doubled" by a cosmic mirror (gravitational lensing).

Key Takeaways:

• The Problem: Scientists couldn't agree on the details of a massive black hole crash because the data looked messy.
• The Fix: They realized the data looked like two signals overlapping. When they modeled it as two signals, the messiness vanished.
• The Reality Check: Two separate crashes happening at once is statistically impossible.
• The Solution: The "two signals" are likely actually one signal that was split and delayed by a massive object in space (gravitational lensing).

This discovery is a big step forward. It shows that when our detectors get better, we might start seeing these "echoes" or "doubles" of cosmic events, and we now have a method to figure out if we are hearing one voice or two.

Technical Summary

Technical Summary of GW231123: Overlapping Gravitational Wave Signals?

Problem Statement
The gravitational wave event GW231123, interpreted as a binary black hole (BBH) merger with a total mass of 190–265 $M_\odot$, represents the heaviest such system detected to date. However, primary parameter estimation analyses of GW231123 have exhibited significant discrepancies in source properties (masses, spins, distance) when using different waveform models (e.g., IMRPhenomXPHM vs. NRSur7dq4). These discrepancies exceed expected statistical fluctuations and systematic differences between models and cannot be reliably reproduced by simulations of isolated signals. The authors investigate whether these inconsistencies arise from the presence of an unaccounted overlapping gravitational wave signal, which could bias parameter estimation if ignored, or from other phenomena such as gravitational lensing or noise artifacts.

Methodology
The authors employ Bayesian model selection to compare two hypotheses: an isolated signal model ($S$) versus a model containing two independent, overlapping signals ($OS$).

• Data Analysis: The analysis utilizes a 4-second data segment centered on the trigger time of GW231123, covering the frequency range [20, 448] Hz.
• Waveform Models: Four distinct waveform models are tested: IMRPhenomXPHM (XPHM), NRSur7dq4 (NRSur), IMRPhenomTPHM (TPHM), and IMRPhenomXO4a (XO4a).
• Bayesian Evidence: Using the Bilby codebase and Dynesty sampler, the authors compute the Bayesian evidence ($Z$) for both models. The log Bayes factor ($\log_{10} \mathcal{B}^S_{OS}$) is calculated to determine model preference, accounting for the penalty of the $OS$ model having twice the number of parameters (Occam's razor).
• Simulations:
  • Recovery Simulations: Two overlapping signals generated by the NRSur waveform were simulated in zero noise and recovered using the isolated signal model with all four waveforms to test if neglecting the second signal reproduces the observed discrepancies.
  • Background Estimation: A set of 100 simulated BBH signals (50 from the general population, 50 from the GW231123 posterior) were analyzed in both simulated Gaussian noise and actual detector noise (excluding the event) to estimate the background distribution of $\log_{10} \mathcal{B}^S_{OS}$. This determines how often noise fluctuations in isolated signals mimic overlapping signal features.
• Prior Odds Calculation: The authors calculate the prior probability of two high-mass BBH signals overlapping within a 0.4-second window, using the merger rate inferred from GW190521 and Poisson statistics.

Key Results

• Model Preference: For all four waveform models, the overlapping signals model ($OS$) is favored over the isolated signal model ($S$). The log Bayes factors range from $\sim 0.2$ to $\sim 4.2$, with the strongest evidence found for the XPHM and NRSur models ($\log_{10} \mathcal{B}^S_{OS} \approx 4.22$ and $1.96$, respectively). These values fall within the top few percent of the background distribution.
• Resolution of Discrepancies: Under the $OS$ model, the significant discrepancies in source properties (mass, spin, distance) observed between the XPHM and NRSur analyses in the $S$ model are largely mitigated. The "louder" signal parameters become consistent across waveforms.
• Signal Characteristics: The recovered "fainter" signal in the $OS$ model exhibits broad posterior distributions due to low signal-to-noise ratio (SNR) but shows parameter estimates (masses, sky location) overlapping with the louder signal. The time delay between the two signals is estimated at approximately 20 milliseconds.
• Simulation Validation: Simulations of overlapping signals recovered with an isolated model successfully reproduced the waveform-dependent discrepancies seen in the GW231123 data, confirming that neglecting an overlapping signal can induce such systematics.
• Background and Noise: Background estimates indicate that while noise artifacts can mimic overlapping signals, the observed foreground statistic for GW231123 is significantly higher than the background for most waveforms. However, the difference in Bayes factors between waveforms (e.g., XPHM vs. NRSur) suggests waveform systematics play a role.
• Comparison with GW190521: Analysis of the previous high-mass event GW190521 shows no significant preference for the overlapping model ($\log_{10} \mathcal{B}^S_{OS} \approx 0$), and the inferred properties of a potential second signal shift toward unphysical high distances, suggesting the overlapping model is not favored for GW190521.
• Prior Odds: The calculated prior odds for two astrophysical high-mass BBH signals overlapping by chance are extremely low ($\sim 10^{-7}$), resulting in a final posterior odds ratio of $\lesssim 10^{-3}$.

Significance and Claims
The paper claims that the overlapping signal model provides a better fit to the GW231123 data than the isolated signal model, effectively resolving the inter-model discrepancies in source property estimation. However, the authors maintain a modest stance regarding the physical interpretation:

• Astrophysical Overlap: The authors explicitly state that the low prior odds of two independent high-mass BBHs overlapping make a coincidental astrophysical overlap highly unlikely.
• Alternative Explanations: The authors propose that the overlapping signal model may serve as a proxy for other phenomena. Specifically, the similarity in properties (mass, sky location) of the two recovered signals, combined with the 20 ms time delay, aligns with the characteristics of a gravitationally lensed signal where two images overlap in time.
• Noise and Systematics: The paper acknowledges that noise artifacts and waveform systematics (particularly in the merger-dominated, high-mass regime) could also produce signatures resembling overlapping signals.
• Conclusion: The work does not definitively confirm an overlapping signal or lensing event but suggests that the overlapping signal hypothesis is a necessary consideration to rule out before confirming gravitational lensing. The findings motivate further investigation into the lensing scenario for GW231123 and highlight the need for methodologies to distinguish between lensing and overlapping signals as detector sensitivity increases.