False Alarm Rates in Detecting Gravitational Wave Lensing from Astrophysical Coincidences: Insights with Model-Independent Technique GLANCE

This study employs the model-independent GLANCE pipeline to quantify the false alarm rates of incorrectly identifying unlensed binary black hole mergers as lensed events, revealing that such misclassifications occur in approximately 0.01% of pairs with time delays exceeding 1000 days under current LIGO sensitivities.

The Gist

The Big Picture: Finding Cosmic Echoes in a Noisy Room

Imagine the universe is a giant, dark concert hall. Every now and then, two massive black holes crash into each other, creating a "chirp" sound known as a Gravitational Wave (GW). Our detectors (like LIGO and Virgo) are like super-sensitive ears trying to hear these whispers.

Sometimes, a massive object (like a galaxy or a giant black hole) sits between the crash and our ears. This object acts like a cosmic magnifying glass (a process called Gravitational Lensing). It bends the space around it, splitting the single "chirp" into multiple copies. These copies arrive at different times, but they sound identical—like hearing the same song play on two different radios in the same room, just with a slight delay.

The Problem: How do we know if we are hearing a "lensed echo" (two copies of the same event) or just two completely different black holes crashing at different times that happen to sound very similar?

This paper is about solving that puzzle. The authors built a tool called GLANCE to find these echoes, but they also wanted to know: How often does GLANCE get tricked by random chance?

---

The Analogy: The "Twin" Confusion

Think of the universe's population of black holes as a massive crowd of people walking through a foggy park.

• The Real Lensing Event: Imagine a person (the black hole) walks past a giant, curved mirror. You see two reflections of the same person walking at the same time. They are perfect twins.
• The False Alarm: Now, imagine two different people in the crowd happen to look exactly alike (same height, same clothes, same walk) and they happen to walk past your window at almost the same time. To a casual observer, you might think, "Wow, those are twins!" But they aren't. They are just two strangers who happen to look similar.

The authors of this paper asked: "If we scan the whole crowd for 'twins', how often will we accidentally pick two strangers who just happen to look alike?"

What They Did: The "GLANCE" Tool

The authors developed a software pipeline called GLANCE (Gravitational Lensing Authenticator using Non-modeled Cross-correlation Exploration).

• How it works: Instead of guessing what the "mirror" (the lens) looks like, GLANCE just listens to the two sounds. It asks: "Do these two waves match perfectly, down to the tiniest detail?"
• The Test: They simulated 3 years of data, creating thousands of fake black hole crashes. They then ran GLANCE over this data to see how many times it would scream, "I found a lensed pair!" when it was actually just two random, similar-looking black holes.

The Results: It's Rare, But Possible

Here is what they found, broken down simply:

1. The "False Alarm" Rate is Tiny:
   Out of all the pairs of black holes that looked similar enough to be confused, GLANCE only falsely identified about 0.01% of them as lensed. That's like finding one fake twin in a crowd of 10,000 people.

   • The Catch: These false alarms mostly happened when the two events were separated by a very long time (over 1,000 days). If two events happen close together in time, it's very unlikely they are just random strangers looking alike.

2. The "Sweet Spot" for Discovery:
   The paper maps out a "safe zone" for finding real lensing.

   • High Magnification + Short Delay: If the "echo" is very loud (high magnification) and arrives very quickly after the first sound (short delay), we can be almost 100% sure it's a real lensed event.
   • Low Magnification + Long Delay: If the echo is quiet and arrives a year later, it's much harder to tell if it's a real echo or just a random coincidence.

3. The Tool is Excellent:
   They tested GLANCE against a "random guesser" and an "ideal detective." GLANCE scored a 99.7% on the scale of perfection (0.997 out of 1.0). It is incredibly good at telling the difference between a real cosmic echo and a random coincidence.

Why This Matters for the Future

We are currently in the "O4" era of gravitational wave detection, where we hear about 200 events a year. But in the future, we will have next-generation detectors that are much more sensitive. They will hear thousands of events.

• The Risk: As the crowd gets bigger, the chance of finding two random "twins" (false alarms) increases.
• The Solution: This paper gives us a rulebook. It tells future scientists: "If you see a lensed pair with a time delay of 5 years, be very careful; it might be a fake. But if you see one with a delay of 1 day and a huge signal, you can be confident it's real."

The Bottom Line

Finding gravitational lensing is like finding a needle in a haystack. This paper proves that the tool we are using (GLANCE) is sharp enough to find the needle without getting stuck on a piece of straw. While there is a tiny risk of being fooled by random chance, the authors show us exactly how to avoid those traps, paving the way for the first confirmed discovery of a "lensed" gravitational wave.

Technical Summary

1. Problem Statement

The detection of gravitationally lensed gravitational waves (GWs) is a critical goal for testing General Relativity and probing the universe's structure. However, confirming a lensed GW event is challenging due to false alarms (False Positives). These false alarms arise primarily from two sources:

• Detector Noise: Instrumental artifacts mimicking signal characteristics.
• Astrophysical Coincidences: Two distinct, unlensed GW events (e.g., merging Binary Black Holes or BBHs) that happen to share similar intrinsic properties (masses, spins) and extrinsic properties (sky position) purely by chance.

Previous work established the GLANCE (Gravitational Lensing Authenticator using Non-modeled Cross-correlation Exploration) pipeline, a model-independent technique to detect lensing. However, it did not quantify the False Alarm Rate (FAR) specifically caused by the astrophysical population of unlensed sources. This paper addresses that gap by estimating the probability that an unlensed pair of BBHs will be incorrectly classified as a lensed pair.

2. Methodology

The authors employed a simulation-based approach to quantify the False Positive Probability (FPP) using the GLANCE pipeline.

• Simulation of GW Population:

  • Used the GWSIM package to generate a population of merging BBHs over a 3-year observation period (simulating LIGO-Virgo-KAGRA O4 sensitivities).
  • Waveforms were generated using the IMRPhenomPv2 model.
  • Events were selected if they had a matched-filter Signal-to-Noise Ratio (SNR) $\geq 8$ in at least two detectors.

• Sky Localization and Pairing:

  • Used BILBY with the DYNESTY nested sampler to estimate sky-localization errors (Right Ascension and Declination) for selected events.
  • Identified sky-overlapping pairs: Event pairs where the 90% credible intervals (C.I.) of their sky positions overlapped. This is a necessary condition for lensing, as lensed images must originate from the same sky patch.

• The GLANCE Technique (Cross-Correlation):

  • Reconstructed GW polarizations ($h_+$ and $h_\times$) from the detector strain data.
  • Performed cross-correlation between the reconstructed polarizations of sky-overlapping event pairs.
  • Calculated a Lensing SNR ($\rho_{\text{lensing}}$) by comparing the cross-correlation signal strength against the noise cross-correlation statistics.
  • The technique relies on the fact that in the Geometric Optics (GO) regime, lensed images have identical frequency, amplitude, and phase evolution, merely shifted in time and magnified.

• False Positive Analysis:

  • Analyzed the distribution of lensing SNRs for the unlensed, sky-overlapping pairs.
  • Generated Receiver Operating Characteristic (ROC) curves by simulating true lensed events (using a Schechter mass distribution for lenses) alongside the unlensed population to calculate True Positive Probability (TPP) vs. False Positive Probability (FPP).

3. Key Contributions

• Quantification of Astrophysical FAR: The paper provides the first model-independent estimate of the FAR specifically due to chance coincidences in the BBH population, separating it from noise-induced false alarms.
• Parameter Space Mapping: The authors mapped the regions in the Chirp Mass vs. Time Delay and Magnification vs. Time Delay parameter spaces where lensing detection is most vulnerable to contamination.
• Performance Validation: Validated GLANCE using ROC curves, demonstrating its ability to distinguish lensed pairs from unlensed coincidences with high fidelity.
• Threshold Definition: Established specific SNR and time-delay thresholds required to claim a confident detection in the current and future detector eras.

4. Key Results

• False Positive Probability (FPP):

  • For a lensing SNR threshold of 1.5, approximately 0.01% of sky-overlapping unlensed event pairs are falsely classified as lensed.
  • No event pairs in the simulated unlensed population exceeded a lensing SNR of 2.0.
  • The FPP decreases significantly as the lensing SNR threshold increases.

• Time Delay Dependence:

  • False alarms are heavily dependent on the time delay ($\Delta t_d$) between events.
  • As the time delay increases (e.g., >1000 days), the probability of chance coincidence from the astrophysical population increases.
  • Confident detections are only feasible in the low time-delay regime (short delays) combined with high magnification, where astrophysical contaminants are statistically unlikely to mimic lensed pairs.

• Mass Distribution:

  • False alarms are most concentrated in the 35–55 $M_\odot$ chirp mass range, which corresponds to the peak of the observed BBH population distribution.
  • High-mass sources ($>20 M_\odot$) contribute more significantly to false alarms than low-mass sources due to their higher detectability rates, despite low-mass sources being more abundant.

• ROC Curve Performance:

  • The GLANCE technique achieved an Area Under the Curve (AUC) of 0.997, indicating performance nearly identical to an ideal lensing detector (AUC = 1.0).
  • True lensed events (simulated) clustered in the low time-delay, high lensing SNR region, well-separated from the unlensed population.

5. Significance and Future Prospects

• Robust Detection Criteria: The study establishes that a lensing detection is robust only if the event pair has a short time delay and high lensing SNR. This provides a clear "safe zone" for future detections to avoid astrophysical false positives.
• Next-Generation Detectors: While future detectors (e.g., Cosmic Explorer, Einstein Telescope) will observe more events, they will also increase the rate of chance coincidences. The authors note that estimating the FAR for these next-generation networks is crucial.
• Methodological Utility: The GLANCE pipeline, being model-independent, is shown to be highly effective at mitigating noise and distinguishing true lensing from population coincidences without relying on specific lens models.
• Implication for GWTC-4 and Beyond: With the growing catalog of GW events (GWTC-4 and beyond), this work provides the statistical framework necessary to claim a confirmed lensed GW detection with high confidence, ensuring that discoveries are not merely statistical flukes of the BBH population.

In summary, this paper demonstrates that while astrophysical coincidences pose a risk for lensing detection, the GLANCE technique can effectively filter them out, provided detections are restricted to high-SNR, short-delay regimes. The false alarm rate due to astrophysical sources is negligible (0.01%) at conservative thresholds, paving the way for confident identification of the first strongly lensed GW events.