🔭 astrophysics

ler: LVK (LIGO-Virgo-KAGRA collaboration) event (compact-binary mergers) rate calculator and simulator

The paper introduces **ler**, a modular and computationally efficient Python package designed to simulate compact-binary merger populations and estimate detectable event rates for current and future LIGO-Virgo-KAGRA networks, including capabilities for modeling both unlensed and strongly lensed systems.

Original authors: Hemantakumar Phurailatpam (Leo), Anupreeta More (Leo), Harsh Narola (Leo), Ng Chung Yin (Leo), Justin Janquart, Chris Van Den Broeck, Otto Akseli Hannuksela, Neha Singh, David Keitel

Published 2026-06-03

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Hemantakumar Phurailatpam (Leo), Anupreeta More (Leo), Harsh Narola (Leo), Ng Chung Yin (Leo), Justin Janquart, Chris Van Den Broeck, Otto Akseli Hannuksela, Neha Singh, David Keitel

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, dark ocean. Occasionally, massive objects like black holes or neutron stars crash into each other, creating ripples in the fabric of space and time called gravitational waves. Since 2015, our "ears" on Earth (the LIGO, Virgo, and KAGRA detectors) have been listening to these ripples.

Most of the time, these ripples travel straight to us, clear and unobstructed. But sometimes, a massive object like a galaxy or a cluster of galaxies sits right in the middle of the path. Think of this galaxy as a giant, invisible magnifying glass (a lens). When the gravitational waves pass through it, the lens splits the signal into multiple copies, makes them louder, and shifts their arrival times. This is called gravitational lensing.

The paper introduces a new software tool called ler (which stands for "LIGO-Virgo-KAGRA Event Rate Calculator"). You can think of ler as a super-powered weather forecast simulator, but instead of predicting rain, it predicts how many of these gravitational wave signals we will "hear" and what they will look like.

Here is how ler works, broken down into simple concepts:

1. The Two Types of Forecasts

Just like a weather model needs to know about clear skies and storms, ler simulates two types of events:

The "Clear Sky" Events (Unlensed): These are the standard signals that travel straight to Earth. ler calculates how often these happen based on how many black holes and neutron stars exist in the universe.
The "Magnified" Events (Lensed): These are the tricky ones where a galaxy acts as a lens. ler simulates the complex math of how a galaxy bends space, splits the signal into multiple images, and amplifies the sound.

2. The "Speedy Chef" Analogy

Calculating these events is usually like trying to cook a massive banquet for a million people by hand—it takes forever and is prone to mistakes.

The Old Way: Scientists used to do these calculations one by one, which was slow and often required huge amounts of computer power.
The ler Way: The authors built ler to be a "speedy chef." They used special cooking techniques (called optimization and parallelization) that allow the software to chop, cook, and plate thousands of simulations at the exact same time.
- The Result: The paper claims ler is up to 1,000 times faster than the old methods. It's the difference between waiting a week for a meal and getting it in seconds.

3. The "Recipe Book" (Modularity)

One of the smartest things about ler is that it's built like a modular kitchen.

If a scientist wants to change the "recipe" (for example, to test a different theory about how black holes form), they don't have to rebuild the whole kitchen. They can just swap out the "ingredient" (the source model) or the "cooking method" (the lens model).
It comes with pre-made recipes for different types of cosmic collisions (black hole pairs, neutron star pairs, etc.) and different types of "lenses" (galaxies), but it also lets scientists add their own custom ingredients.

4. Why Do We Need This?

Scientists need ler for three main reasons, according to the paper:

To Count the Stars: It helps them figure out the true rate of cosmic collisions in the universe, correcting for the fact that our detectors can only "hear" signals that are loud enough.
To Find the Hidden Gems: It helps them predict how many "magnified" signals we should expect to find. If we find a signal that looks like it was lensed, ler helps verify if it's real or just a fluke.
To Plan for the Future: It helps scientists forecast what our detectors will see in the future, helping them prepare for new observatories.

In Summary

ler is a fast, flexible, and powerful computer program that helps scientists simulate the "noise" and "signals" of the universe. It acts as a bridge between the messy reality of the cosmos and the clean data our detectors collect, allowing researchers to understand how often black holes collide and how often the universe's own "magnifying glasses" help us hear them.

Note: The paper explicitly states that no generative AI tools were used to write the software or the manuscript.

Technical Summary of ler: LVK Event Rate Calculator and Simulator

Problem Statement
The LIGO-Virgo-KAGRA (LVK) collaboration routinely detects gravitational waves (GWs) from compact-binary mergers. While most signals are unlensed, a fraction undergoes strong gravitational lensing by intervening massive objects (e.g., galaxies), resulting in magnified, time-shifted, and phase-shifted multiple images. Accurately modeling both unlensed and lensed populations is critical for astrophysical interpretation, inferring merger rate histories, and validating future lensing candidates. However, existing tools present limitations: general-purpose inference packages (e.g., bilby) and rate tools (e.g., gwcosmo) lack a unified, configuration-driven pipeline to jointly generate large-scale unlensed and lensed populations while propagating lensing image properties into detectability. Conversely, dedicated lensing tools (e.g., lenstronomy) are not optimized for high-throughput population-level simulation. Furthermore, previous studies often relied on limited sample sizes, lacked convergence diagnostics, and neglected the propagation of population-model hyperparameter uncertainties into rate predictions.

Methodology
The paper introduces ler, a statistics-based Python framework designed to compute detectable rates for both unlensed and strongly lensed GW events through an end-to-end Monte Carlo simulation pipeline. The methodology is structured around two primary rate calculations:

Unlensed Rates: The pipeline draws compact-binary source populations from astrophysical priors, evaluates their detectability, and integrates over detected events.
Lensed Rates: This extends the flow by sampling lens-galaxy and source-redshift parameters under strong-lensing conditions. It solves the lens equation to derive image properties, applies image-level detectability criteria, and integrates over detectable lensed events.

Key Technical Features:

Modularity: The software separates scientific steps from computational machinery, allowing users to swap source models (e.g., BBH, BNS, NSBH), lens models (e.g., SIS, SIE, EPL+Shear), and detection criteria without rewriting the core pipeline.
Computational Optimization: To address the high computational cost of large-scale sampling and lens equation solving, ler employs:
- Just-in-Time (JIT) Compilation: Utilizing numba to compile repeatedly evaluated parts of the pipeline.
- Parallelization: Using Python multiprocessing and multithreading for Monte Carlo loops.
- Efficient Sampling: Implementing inverse-transform and importance sampling to avoid the inefficiencies of rejection sampling.
- Backend Integration: It optionally uses lenstronomy for lensing calculations but includes a custom, compiled analytical solver for EPL+Shear models and caustic/cross-section calculations to maximize speed.
Detection Probability ( $P_{det}$ ): The framework integrates with gwsnr for efficient detection-probability calculations, which are used to construct selection functions for hierarchical population inference.

Results and Performance
The paper reports significant performance gains achieved through the optimization strategies described above:

Speedup: Benchmarks indicate speedups of approximately 180× for unlensed populations and 5,189× for lensed populations compared to conventional methods that disable JIT compilation and parallelism.
Execution Time: On a standard six-core processor (Apple M2 Pro, 16GB RAM), a batch of $10^5$ samples takes approximately 300 ms for unlensed calculations and 30 s for lensed calculations.
Convergence: The tool provides convergence statistics from repeated batches, ensuring that rate estimates are stable and that accuracy is controlled by sample size rather than closed-form approximations.
Overhead: Initial JIT compilation introduces a one-time overhead of ~2 seconds for unlensed batches and ~25 seconds for lensed batches (including image-property calculations), after which per-sample costs are significantly reduced.

Significance and Impact
The paper positions ler as a robust framework for high-level population studies and astrophysical forecasting within the GW community. Its significance is demonstrated through:

Community Adoption: The package has recorded over 3,000 downloads from PyPI and is actively used by researchers ranging from students to faculty across multiple institutions.
Scientific Utility: It has been integrated into workflows for validating lensing candidates (analyzing time delays and magnification ratios), forecasting sub-threshold lensed event detections for future observatories, generating lensed-event statistics, and calculating posterior odds.
Reproducibility: By providing a unified, configuration-driven pipeline, ler addresses gaps in previous studies regarding sample size convergence and uncertainty propagation, supporting reproducible research in LVK analyses and multimessenger applications (e.g., strongly lensed gamma-ray bursts).

The software is open-source, peer-reviewed within the LVK collaboration, and designed to be adaptable for future detector networks and multimessenger studies.