Basilic: An end-to-end pipeline for Bayesian burst inference and model classification in gravitational-wave data

This paper introduces Basilic, a modular, user-friendly pipeline built on the bilby framework for Bayesian inference and model selection of gravitational-wave bursts, demonstrating its capabilities through a study of degeneracies between binary black hole mergers and cosmic string signals.

The Gist

Imagine the universe is a giant, noisy radio station. For over a decade, the LIGO, Virgo, and KAGRA detectors have been listening for specific "songs" from space: the dramatic, spiraling duets of black holes colliding. We know how to recognize these songs because they have a distinct, rising pitch (a "chirp").

But what about the static? What about the weird, short bursts of noise that don't fit any known song? These are called gravitational-wave bursts. They could be from exploding stars, cosmic strings (like tiny, vibrating cracks in the fabric of space), or even glitches in our own detectors. The problem is, in the middle of a storm of static, it's incredibly hard to tell if a short "pop" is a real cosmic event or just a random glitch.

This paper introduces Basilic, a new, super-smart tool designed to solve this mystery. Here is a simple breakdown of what it does and what the scientists found.

1. What is Basilic? (The "Universal Translator")

Before Basilic, analyzing these weird bursts was like trying to solve a puzzle where every piece was a different shape, and you had to build a new table for every single piece. It was slow, messy, and required a PhD in computer science just to get started.

Basilic is like an automated, all-in-one puzzle factory.

• It's a Pipeline: You give it a single instruction (a configuration file), and it does everything: it grabs the data, cleans the noise, tests different theories, and spits out a clear report.
• It's Modular: Think of it like a video game character with a backpack. You can swap out the "models" (the theories of what caused the burst) easily. Maybe you want to test if it was a supernova? Swap the backpack. Maybe a cosmic string? Swap it again.
• It's Fast: It uses a massive network of computers (HTCondor) to run thousands of simulations at once, so scientists don't have to wait weeks for results.

2. The Big Mystery: The "Cosmic String" vs. "Black Hole" Mix-up

The scientists used Basilic to investigate a specific confusing situation.

Imagine you hear a short, sharp crack in the distance.

• Theory A: It was a giant, heavy black hole merging (a "heavy" event).
• Theory B: It was a cosmic string snapping (a theoretical, exotic event).

Usually, these two sounds are very different. A black hole merger is a long, rising chirp. A cosmic string is a sharp, sudden burst. However, if the black hole is extremely heavy, the "chirp" happens so fast that it looks just like a sharp crack. It's like hearing a distant thunderclap and not knowing if it was a giant storm or a small firecracker.

3. The Experiment: The "Cosmic Soundboard"

To figure out when this mix-up happens, the team didn't just wait for real events. They created a massive simulation campaign.

• They took a "soundboard" of 10 different types of heavy black hole mergers.
• They played each one 100 times, but each time they added a different layer of "static" (simulated detector noise).
• They asked Basilic: "Is this a black hole or a cosmic string?"

4. The Surprising Discovery

The results were fascinating. They found that the confusion isn't just about how heavy the black holes are.

• The Weight Factor: Yes, very heavy black holes look like cosmic strings.
• The Spin Factor (The New Discovery): They found that even if the black holes aren't super heavy, if they are spinning in the wrong direction (anti-aligned), they also start to sound like cosmic strings!

The Analogy: Imagine two people shouting.

• If they are both very loud (high mass), their voices blend into a roar that sounds like a jet engine (cosmic string).
• But, if they are shouting while spinning around wildly in opposite directions (anti-aligned spins), their voices also get distorted into a weird, staticky roar, even if they aren't that loud to begin with.

5. What Do We Do When We're Not Sure?

The paper concludes with a smart strategy for when the computer says, "I'm 50/50 on this."
Instead of just guessing, they propose a two-step detective check:

1. The "Reality Check" (Posterior Predictive Checks): "If this theory were true, would the data look like this?" If the answer is "No, the data looks too weird for this theory," we discard it.
2. The "Shape Match" (Waveform Match): "Do the shapes of the waves actually overlap?" If the black hole wave and the cosmic string wave are mathematically identical in the region where the data is, then we admit: We can't tell them apart yet. We need better data (less noise) to solve the puzzle.

The Bottom Line

Basilic is a game-changer because it turns a difficult, manual science project into a routine, automated process. It helps us understand that in the noisy universe, heavy black holes and exotic cosmic strings can sound exactly the same.

The authors are essentially saying: "We built a better microscope (Basilic) to look at these short bursts. We found that sometimes, the universe plays tricks on us, making heavy black holes look like cosmic strings. But now, we have a checklist to know when we're looking at a trick and when we've actually found something new."

Technical Summary

1. Problem Statement

Gravitational-wave (GW) astronomy has successfully detected compact binary coalescences (CBCs) for over a decade. However, short-duration transient signals ("bursts") from sources like core-collapse supernovae, cosmic strings, or unknown phenomena remain observationally unconfirmed.

• The Challenge: Current burst detection pipelines (e.g., coherent WaveBurst) are optimized for rapid, unmodeled discovery using frequentist statistics. They lack the capability to perform rigorous Bayesian model selection to distinguish between competing physical hypotheses (e.g., a binary black hole merger vs. a cosmic string cusp).
• The Degeneracy Issue: Short-duration signals often have low signal-to-noise ratios (SNR) and few observed cycles. This leads to morphological degeneracy, where physically distinct models (like high-mass CBCs and cosmic strings) produce nearly identical detector-frame waveforms, making classification difficult.
• The Software Gap: While the bilby library provides Bayesian inference tools for CBCs, there is no standardized, end-to-end pipeline for burst analysis that handles automated setup, scalable injection campaigns, and post-processing for model comparison.

2. Methodology: The Basilic Pipeline

The authors introduce Basilic (BAyesian tranSIent modeLIng and Classification), a dedicated pipeline built on top of the bilby framework and GWpy.

• Architecture & Workflow:

  • Modularity: Built on bilby for likelihood computation and nested sampling (using dynesty), ensuring robust Bayesian inference.
  • Automation: Uses a single YAML configuration file to define data sources, Power Spectral Densities (PSDs), priors, and hypotheses.
  • Scalability: Integrates with HTCondor for distributed computing, enabling large-scale injection campaigns and single-event analyses without manual scripting.
  • Post-Processing: Automatically generates summary reports, including Bayes factor tables, parameter posteriors, and diagnostic plots.

• Model Catalog:
  Basilic includes a growing registry of parametric burst models, extending beyond standard CBC approximants:

  • Core-Collapse Supernovae (CCSN): Asymmetric Gaussian chirplets and three-peak templates.
  • Cosmic Strings: Cusp, kink, and kink-kink interactions (modeled as power laws).
  • Memory Effects: Linear gravitational memory (slow-rise, broadband).
  • Generic Models: Power laws and damped sinusoids (useful for glitch mitigation).

• Advanced Diagnostic Strategy:
  To address inconclusive Bayes factors, the authors propose a two-stage diagnostic framework implemented in Basilic:

  1. Posterior Predictive Checks (PPC): Tests model adequacy by generating replicated datasets from the posterior. If a model cannot reproduce the observed data (low Bayesian p-value), it is rejected.
  2. Morphological Match Analysis: If PPCs are inconclusive, the pipeline calculates the waveform match (overlap) between the posterior distributions of competing models. High overlap indicates intrinsic morphological degeneracy rather than just noise.

3. Key Contributions

1. Development of Basilic: The first end-to-end, user-friendly pipeline specifically designed for Bayesian burst inference and model classification in the LVK (LIGO-Virgo-KAGRA) context.
2. Systematic Injection Campaign: A controlled study injecting high-mass Binary Black Hole (BBH) signals into Gaussian noise to test their confusion with Cosmic String (CS) cusp waveforms.
3. Discovery of New Degeneracies:
   • Confirmed the known degeneracy between high-mass BBHs and cosmic strings.
   • Novel Finding: Discovered that high anti-aligned component spins (even at moderate masses) can induce a similar degeneracy, causing BBH signals to mimic cosmic string waveforms in low-SNR regimes.
4. Operational Framework: Established a rigorous protocol for handling inconclusive model selection using PPCs and waveform matching, moving beyond simple Bayes factor thresholds.

4. Key Results

The study utilized a network of LIGO Hanford (H1) and Livingston (L1) with a fixed optimal SNR of $\rho_{opt} = 6$ (a low-SNR regime typical for burst candidates).

• Mass Dependence: As the total mass of the BBH increases, the correlation between the Bayes factors for the BBH hypothesis and the Cosmic String hypothesis strengthens. High-mass BBHs are significantly harder to distinguish from cosmic strings.
• Spin Alignment Dependence:
  • Aligned spins: Showed weaker correlation with cosmic string models.
  • Anti-aligned spins: Showed a strong correlation. High-magnitude, anti-aligned spins result in waveforms that are morphologically similar to cosmic string cusps, leading to ambiguous model selection.
• Mass Ratio: Variations in mass ratio ($q$) did not significantly drive model confusion compared to total mass and spin alignment.
• Noise Impact: In the absence of noise, BBH and CS waveforms are distinct. The degeneracy is noise-enabled; specific noise realizations at low SNR obscure the "chirp" signature of the BBH, making it indistinguishable from a burst.
• Quantitative Metrics: The study measured the correlation coefficient ($r$) of the Bayes factors. For example, a low-mass, aligned-spin case had $r \approx 0.39$, while a high-mass, anti-aligned-spin case had $r \approx 0.87$, quantifying the increased ambiguity.

5. Significance and Future Implications

• Interpretation of Future Events: As LVK sensitivity improves, more short-duration signals will be detected. This work provides the tools to correctly interpret them, preventing misclassification of high-mass/anti-aligned BBHs as exotic cosmic string events (or vice versa).
• Robustness in Low-SNR Regimes: The paper highlights that an inconclusive Bayes factor does not necessarily mean the models are physically identical; it may simply mean the SNR is too low to break the degeneracy. The proposed PPC and match-based diagnostics allow researchers to distinguish between "noise-limited" ambiguity and "intrinsic" morphological overlap.
• Standardization: Basilic lowers the technical barrier for burst analysis, allowing for reproducible, large-scale injection campaigns that were previously impractical with ad-hoc scripts.
• Scientific Rigor: By formalizing the approach to model degeneracy, the paper sets a new standard for how the GW community should handle sub-threshold or ambiguous transient events, ensuring that claims of exotic physics are backed by robust statistical evidence.

In summary, Basilic bridges the gap between burst detection and rigorous Bayesian inference, providing both the software infrastructure and the methodological framework necessary to navigate the complex landscape of short-duration gravitational-wave signals.