⚛️ general relativity

How general relativity stacks up: Testing gravity by coherently combining gravitational-wave observations

This paper proposes and applies a framework to coherently combine gravitational-wave observations from 28 binary black hole mergers to test general relativity in the high-curvature regime, finding no evidence for deviations while demonstrating how such a population-level analysis can distinguish genuine physics beyond GR from systematic waveform errors.

Original authors: Lachlan Passenger, Nir Guttman, Paul D. Lasky, Eric Thrane

Published 2026-10-22

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Lachlan Passenger, Nir Guttman, Paul D. Lasky, Eric Thrane

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

Einstein's theory of general relativity describes gravity not as a force, but as a curvature of space and time caused by mass. For over a century, this theory has passed every test we have thrown at it, from the orbit of planets to the bending of starlight. However, the most extreme environments in the universe—where black holes collide and merge—offer a unique laboratory to see if the theory holds up under the most intense pressure imaginable. When two black holes spiral toward each other, they ripple the fabric of spacetime itself, sending out gravitational waves. By listening to these waves, scientists can check if the behavior of the merging objects matches the precise predictions of Einstein's equations. If the waves differ even slightly from what the theory says they should be, it could signal the existence of new physics, a hidden layer of reality that general relativity has not yet captured.

A team of researchers has developed a new method to listen for these subtle discrepancies by combining the signals from dozens of black hole collisions. While individual collisions are often too faint to reveal tiny deviations, the researchers realized that if a deviation exists, it should follow a predictable pattern across many events. They proposed that any departure from Einstein's theory would likely be strongest at the moment the black holes smash together, and that the nature of this departure would change smoothly depending on the total mass of the colliding pair. To test this, they created a conceptual map that charts how the gravitational waves should look if they deviate from the standard theory, organizing this map by the mass of the black holes involved.

The team applied this framework to data from twenty-eight high-quality black hole mergers recorded by the LIGO, Virgo, and KAGRA detectors. These events were selected because their signals were loud and clear enough to allow for a detailed analysis. The researchers first subtracted the perfect general relativity prediction from the actual data, leaving behind only the "residual" noise. They then looked for any coherent pattern in this leftover signal that matched their map of potential deviations. The result was a clean slate: the data showed no evidence of any deviation from Einstein's theory. The leftover signals were consistent with random noise, suggesting that general relativity remains unbroken even in the most violent cosmic collisions.

A crucial strength of this new approach is its ability to distinguish between a genuine discovery and a flaw in the scientists' own tools. If the researchers had found a deviation that looked the same regardless of the black holes' masses, it would have suggested a systematic error in the computer models used to generate the theoretical waveforms. However, because their method specifically looked for deviations that change with mass, they could confidently rule out such errors. The fact that they found nothing, even after searching for a pattern that would shift with mass, reinforces the conclusion that the theory is correct. The researchers also tested their method using simulated data where they knew a fake deviation had been inserted; in those cases, their framework successfully recovered the signal, proving the method works as intended.

This work represents a significant step forward in how scientists analyze the growing catalog of gravitational wave events. Instead of treating each collision as an isolated mystery, the team demonstrated how to combine information from many events to build a population-level test. By assuming that deviations would behave in a specific, smooth way across different masses, they created a powerful filter that can separate true new physics from the imperfections of our current models. While this specific study found no cracks in Einstein's theory, the framework they built provides a robust tool for future searches. As more black hole mergers are detected, this method will allow scientists to probe the limits of gravity with increasing precision, ensuring that if a deviation does exist, it will be found.

Technical Summary: "How general relativity stacks up: Testing gravity by coherently combining gravitational-wave observations"

Problem Statement
Testing General Relativity (GR) in the strong-field, highly dynamical regime is a primary science goal of gravitational-wave astronomy. While individual event analyses (e.g., ringdown tests, inspiral-merger-ringdown consistency checks) have confirmed GR, they suffer from limited resolving power driven by signal-to-noise ratio (SNR) and fail to characterize how potential deviations might vary across a population of mergers. Existing population-level approaches face a dichotomy: highly parameterized models (e.g., Maselli et al. 2024) may lack generality, while stochastic models (e.g., Isi et al. 2019a; Dideron et al. 2026) treat deviations as random draws, failing to enforce the physical expectation that identical binaries should exhibit identical deviations. Furthermore, distinguishing genuine new physics from waveform systematics (which are scale-invariant) remains a significant challenge.

Methodology
The authors propose a framework to coherently combine gravitational-wave measurements to construct a "deviated-strain map," $s_{+,\times}(\tau, M)$ , representing the non-GR contribution to the strain as a function of natural time $\tau$ and source-frame total mass $M$ .

Signal Model: The framework assumes deviations are (i) strongest near the merger, (ii) vary smoothly with total mass, and (iii) are most prominent in the $\ell=|m|=2$ mode. The deviated strain is modeled as a zero-mean Gaussian Process (GP) in natural time, with hyper-parameters ( $\Lambda$ ) controlling amplitude, characteristic frequency, duration, and coherence time.
Two-Step Pragmatic Approach: A full Bayesian treatment of the joint posterior for the map parameters and hyper-parameters is computationally intractable due to the high dimensionality ( $\sim 8 \times 10^5$ $\sim 8 \times 1 0^{5}$ parameters) and the need to invert massive covariance matrices ( $O(10^{10})$ $O (1 0^{10})$ elements). To circumvent this, the authors employ a two-step frequentist approximation:
1. Single-Event Analysis: Assuming independence between events ( $\Delta M \to 0$ ), they estimate the GP hyper-parameters $\Lambda$ by maximizing the likelihood of the residual data (after subtracting the best-fit GR waveform) for each event individually.
2. Coherent Combination: Using the estimated $\Lambda$ , they reconstruct the deviated strain for each event on a common natural-time grid. They then reintroduce mass correlations by weighting events based on their total mass difference using a kernel $w_{ij}(\Delta M) = \exp[-(M_i - M_j)^2 / (2\Delta M^2)]$ . This allows the construction of a population-level map where events with identical mass yield identical strain, and strain decorrelates over a scale $\Delta M$ .
Systematics Discrimination: The framework leverages the scale-invariance of GR. Waveform systematics manifest as mass-independent (vertical) bands in the deviated-strain map ( $\Delta M \to \infty$ ), whereas genuine deviations from GR are expected to exhibit mass dependence.

Key Contributions

Deviated-Strain Map: Introduction of a phenomenological "map" that visualizes deviations as a function of total mass and time, moving beyond single-event or purely stochastic population models.
Coherent Combination Framework: A computationally feasible method to combine data from multiple events to boost the significance of potential GR violations, utilizing Gaussian Processes to enforce smoothness in mass and time.
Systematics vs. New Physics: A mechanism to distinguish between waveform modeling errors and new physics by analyzing the mass-dependence of the recovered signal. If the signal favors $\Delta M \to \infty$ , it suggests systematics; if it favors finite $\Delta M$ , it suggests new physics.
Validation: The framework is validated using synthetic data (200 events) where a known deviated strain signal is injected. The method successfully recovers the qualitative shape of the injected map and yields a statistically significant detection statistic ( $\ln \hat{\lambda}_{pop} = 9.49$ ).

Results
The authors applied the framework to 28 high-SNR binary black hole mergers from the LIGO-Virgo-KAGRA Fifth Gravitational-Wave Transient Catalog (GWTC-5).

Selection Criteria: Events were selected based on network SNR $\gtrsim 15$ , dominant ringdown frequencies within [50 Hz, 511 Hz], and source-frame total mass $< 120 M_\odot$ .
Analysis: After subtracting maximum-likelihood GR waveforms (IMRPhenomXPHM or IMRPhenomXPHM-SpinTaylor), the residual data were analyzed.
Outcome: The analysis found no evidence for deviations from GR. The likelihood ratio comparing the hypothesis of a non-zero deviated strain ( $\hat{s} > 0$ ) to the null hypothesis ( $\hat{s} = 0$ ) yielded $\ln \hat{\lambda}_{pop} = 0.46$ , corresponding to a p-value of $0.05$. The reconstructed deviated-strain maps showed no statistically significant structure.

Significance and Claims
The paper positions this work as a "proof of concept" and a "plausible starting point" for a more general framework. The authors explicitly acknowledge that their approach relies on several approximations (e.g., ignoring waveform subtraction uncertainty, total mass uncertainty, and viewing angle uncertainty, and treating the problem as frequentist rather than fully Bayesian) which render the solution "sub-optimal" and "inelegant" compared to a full generative model.

However, the significance lies in the demonstration that:

Coherent combination of events can reconstruct a mass-dependent deviated-strain map.
The mass-dependence of the signal provides a powerful discriminator between waveform systematics and genuine new physics.
Despite the approximations, the method can qualitatively recover injected signals in synthetic data.

The authors conclude that while no deviations were found in GWTC-5, the framework establishes a methodology to rigorously distinguish scale-invariant waveform errors from scale-dependent physics beyond GR in future, larger catalogs. They outline a path forward involving dimensional reduction (principal component analysis) to enable a fully Bayesian implementation that avoids the current approximations.

More like this