⚛️ general relativity

Search for strong lensing of gravitational waves in the binary black hole events from O1-O4a

This paper presents a comprehensive search for strongly lensed binary black hole gravitational wave events across the O1–O4a observing runs using an enhanced Bayesian pipeline and simulations, finding no evidence of lensing to establish a 90% upper bound of 1.4% on the lensing fraction while highlighting the importance of cross-catalog analysis and forecasting future detection probabilities.

Original authors: Ankur Barsode, Koustav N. Maity, Parameswaran Ajith

Published 2026-07-10

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Ankur Barsode, Koustav N. Maity, Parameswaran Ajith

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the universe as a giant, cosmic game of "Whac-A-Mole." Every time two black holes smash together, they send out a ripple in space-time called a gravitational wave. Usually, we catch just one ripple. But sometimes, a massive object like a galaxy or a cluster of galaxies sits right in the path of that ripple, acting like a giant, invisible magnifying glass. This is called gravitational lensing.

If lensing happens, that single ripple gets bent and split, arriving at our detectors as multiple copies of the same signal—like seeing a reflection of yourself in a funhouse mirror, but the "mirror" is a galaxy billions of light-years away. These copies might arrive a few hours or days apart, and they might be louder (magnified) or slightly shifted in time, but they are fundamentally the same event.

The Big Hunt
In this paper, a team of scientists played detective with data from the LIGO, Virgo, and KAGRA detectors. They looked at 207 unique black hole collisions detected during the first four observing runs (O1 through O4a). They didn't just look at the official list; they also checked lists from other groups (like IAS and OGC) to make sure they didn't miss any heavy-hitting candidates.

Their goal? To find pairs of signals that look like twins. They used a super-fast computer pipeline called Posterior Overlap 2.0 (PO2.0). Think of this pipeline as a high-tech matchmaker. It takes two signals and asks: "Are you the same event seen twice, or just two random events that happen to look a bit alike?"

The Verdict: No Twins Found
After checking 21,321 possible pairs of signals, the answer was a resounding "no."

The Finding: The team found zero evidence of strong lensing.
The Confidence: They calculated that for their top candidate pair, the chance of it actually being a lensed event is less than 0.6%. That's like flipping a coin and getting heads less than once in every 160 tries.
The Ruling Out: They explicitly ruled out the idea that the "top" candidates found in previous searches were actually lensed. Even the pairs that other scientists thought looked interesting turned out to be just random coincidences when tested with this new, more rigorous method.

Why Did They Look So Hard?
You might wonder, "If they didn't find anything, why bother?"

The "False Alarm" Problem: Sometimes, two totally unrelated black hole crashes can look similar just by accident. The scientists wanted to make sure they weren't fooling themselves. They ran massive simulations to see how often these "accidental matches" happen. They found that the top matches in their data were just as likely to be accidents as they were to be real lensed events.
The "Heavy" Candidates: They noticed a pattern: the pairs that looked most like lensed events were usually very massive black holes. This makes sense because heavy black holes are more likely to be lensed. However, even these heavyweights didn't pass the test.
The "IAS" Factor: Interestingly, 5 out of the top 9 most promising pairs came from the IAS catalog (a group outside the main collaboration). This highlights that looking at data from multiple groups is crucial, because different teams might spot different things.

What Does This Mean for the Future?
Even though they didn't find a lensed pair, the search wasn't a failure. It set a strict limit: if lensing is happening, it's happening in less than 1.4% of all black hole collisions (with 90% confidence).

The paper also looked into a crystal ball to predict when we might finally catch a lensed pair. Based on their simulations:

By the end of the current run (O4), there's about a 20% chance of finding one.
By an intermediate run (IR1), that jumps to 23%.
By the fifth run (O5), the odds skyrocket to 67%.

The Bottom Line
Right now, the universe hasn't handed us a lensed gravitational wave yet. But the scientists are confident that with better detectors and more data coming in the next few years, we will eventually catch a glimpse of these cosmic reflections. Until then, the "no lensing found" result helps them rule out some wild theories about how often black holes merge and what dark matter might be made of. The hunt is on, and the next few years are expected to be very exciting!

Technical Summary: Search for Strong Lensing of Gravitational Waves in Binary Black Hole Events from O1–O4a

Problem Statement
Strong gravitational lensing of gravitational waves (GWs) by intervening galaxies and galaxy clusters is predicted to produce multiple, time-delayed copies of the same signal. While such lensing offers profound opportunities for astrophysics, cosmology, and testing General Relativity, no confirmed detection has been made to date. Previous searches by the LIGO-Virgo-KAGRA (LVK) Collaboration and external groups have been limited by computational constraints. Specifically, optimal Bayesian model selection (calculating the full Bayes factor for all event pairs) is computationally prohibitive, forcing prior studies to use sub-optimal detection statistics in a first stage. Furthermore, previous searches often ignored pairs spanning different observing runs and excluded candidates identified by non-LVK catalogs (e.g., IAS, OGC), potentially missing high-mass events where lensing is more probable. Additionally, the statistical significance of null results has been difficult to quantify due to the high cost of running extensive background simulations.

Methodology
This study performs a comprehensive search for strongly lensed binary black hole (BBH) pairs across the first three observing runs (O1–O3) and the first part of the fourth run (O4a). The analysis integrates events from the LVK, the Open Gravitational-wave Catalog (OGC), and the Independent Analysis of Signals (IAS) catalogs, totaling 207 unique events and 21,321 possible pairs.

The core of the methodology is the Posterior Overlap 2.0 (PO2.0) pipeline. This Bayesian model-selection tool reformulates the Bayes factor ( $B^L_U$ ) as an inner product of the posterior distributions of two individual events, weighted by population priors. This approach achieves a computational speedup of $\sim 10^4$ compared to traditional joint parameter estimation, allowing for the analysis of all pairs without sub-optimal pre-filtering.

Key methodological components include:

Astrophysical Priors: The search assumes intrinsic BBH parameters follow the GWTC-4 catalog, extrapolated to high redshifts using the Dominik et al. (2013) merger rate model. Lens populations are modeled as Singular Isothermal Ellipsoids (SIE) with masses ranging from $10^8$ to $10^{15} M_\odot$ , based on the Halo Mass Function. This prior is considered "optimistic" toward lensing, as it includes high-mass halos that favor larger time delays and magnifications.
Significance Metrics: The authors define new metrics to assess statistical significance:
- False Positive Probability (FPP): The probability that at least one unlensed pair in the catalog mimics a lensed signal.
- Probability of Lensing ( $p_L$ ): The posterior probability that a specific pair is lensed.
- Catalog Purity ( $c_L$ ): The fraction of truly lensed pairs in a catalog defined by a specific Bayes factor threshold, assessing population-level detection.
Simulations: Realistic background (unlensed) and foreground (lensed) simulations are conducted to characterize the pipeline's efficiency and the statistical significance of candidates.

Key Results

No Evidence for Lensing: The search found no statistically significant evidence for strong lensing. The top-ranked candidate pair (GW200210_035448 and GW200210_100022) has a posterior odds of $10^{-1.89}$ and a probability of being lensed ( $p_L$ ) of less than 0.6%.
Top Candidates: The top 5 candidate pairs all exhibit moderate to high detector-frame total masses ( $120-450 M_\odot$ ) and time delays ranging from hours to days. Notably, 5 of the 10 unique events in the top 5 pairs originate from the IAS catalog, highlighting the importance of including external catalogs that are more sensitive to high-mass events.
Constraints on Lensing Fraction: Based on the non-detection, the authors place a 90% upper bound on the strong lensing fraction at 1.4%. This constraint was derived using a mixture model approach that accounts for the uncertainty in identifying individual pairs, rather than assuming all events are certainly unlensed.
Re-evaluation of Previous Candidates: Previously identified candidate pairs from earlier searches (e.g., Hannuksela et al. 2019; Abbott et al. 2021b) were re-analyzed. All were ranked 15th or lower in this study, exhibiting low scores across all significance metrics, confirming they are likely unlensed.

Significance and Claims
The paper claims to address critical limitations of previous lensing searches by:

Computational Efficiency: Utilizing PO2.0 to perform a near-optimal Bayesian search over the entire dataset without sub-optimal screening.
Comprehensive Scope: Including events from multiple catalogs (LVK, IAS, OGC) and allowing for pairs formed across different observing runs.
Robust Statistics: Providing rigorous background simulations to quantify the statistical significance of null results and defining population-level metrics (purity) to assess the utility of candidate catalogs.

The authors conclude that while no lensed events have been detected in the O1–O4a data, the probability of a $3\sigma$ detection is forecasted to be $\sim 20\%$ by the end of O4, $\sim 23\%$ by the intermediate run (IR1), and $\sim 67\%$ by the fifth observing run (O5). The work emphasizes that the non-detection itself is valuable for ruling out extreme models of BBH merger rates and formation channels. The inclusion of IAS events is highlighted as a significant improvement, as these pipelines recover high-mass events that are more likely to be lensed, though the authors note that combining catalogs introduces selection effect complexities that require future investigation.

More like this