arXiv subjects

Eric Thrane

Publications and source records attributed to Eric Thrane.

At least 19 recordsLinked to original sources

arXiv2026

$\texttt{BilbyFlow}$: user-friendly neural posterior estimation for gravitational-wave astronomy

Bayesian inference plays a central role in the new field of gravitational-wave astronomy. However, traditional Bayesian inference with stochastic samplers is computationally expensive, taking hours to days per event. Transformative changes are therefore required to enable the science of next-generation observatories whose event rates and signal-to-noise ratios will increase significantly over the current generation. Recent work has shown that neural posterior estimation (NPE) is a promising path forward. A neural net is trained to approximate the posterior distribution of gravitational-wave parameters, allowing generation of posterior samples in a fraction of the time required by stochastic samplers. In this work, we introduce $\texttt{BilbyFlow}$, which harnesses the power of NPE in the popular $\texttt{Bilby}$ code suite. We use $\texttt{BilbyFlow}$ to analyze a subset of 38 high-mass events from the third LIGO-Virgo-KAGRA Gravitational-Wave Transient Catalog (GWTC-3). For 29 events (76\%), we obtained an importance-sampling efficiency $>$1%, allowing us to produce reliable posterior distributions within 3 min - 1.5 hours. For the other events, with importance-sampling efficiency $\ll$1%, the run time can be as long as 35 hours. We achieve a median importance-sampling efficiency of 7%, which is roughly comparable to the $\texttt{DINGO}$ package. We aim to significantly improve this efficiency with further development to make the runtime more reliably $O(\text{min})$. $\texttt{BilbyFlow}$ is open source and $\texttt{pip}$-installable.

astro-ph.IM↗

arXiv2026

The properties of the first continuous gravitational waves: Optimizing pulsar timing observations for supermassive black hole binary detection

Several pulsar timing arrays have reported evidence for a nanohertz gravitational-wave background. This background is thought to arise from the superposition of signals from a population of inspiraling supermassive black hole binaries. A key test of this interpretation is the search for individually resolvable binaries whose signals emerge above this background. Here we investigate this possibility by addressing two related questions: (1) what are the most probable properties of the first resolvable binary; and (2) how can pulsar timing observations be optimized to maximize its prospects for electromagnetic identification? By simulating populations of supermassive black hole binaries constrained by evidence for the gravitational-wave background and studying the loudest binary in each simulation, we infer the expected properties of the first individually resolvable systems. If MeerKAT continues its current observing strategy for a total of 9 years, the probability of detecting an individual binary is $\approx10-27\%$. Improving the timing precision by a factor of two changes this probability to $\approx11-28\%$, whereas increasing the cadence by a factor of four gives $\approx12-30\%$. Combining both improvements yields a detection probability of $\approx12-33\%$. Furthermore, implementing a high-cadence observing campaign preferentially resolves higher-frequency systems ($\gtrsim\unit[10]{nHz}$), whose shorter orbital periods make them considerably easier to identify through electromagnetic observations than lower-frequency binaries that repeat infrequently. We discuss prospects of electromagnetic follow-up of gravitational-wave resolved binaries with current high-cadence all-sky optical surveys.

astro-ph.HE↗

arXiv2026

Revealing Four Subpopulations of Binary Black-Hole Mergers with the Fifth Gravitational-Wave Transient Catalog

Gravitational-wave data are beginning to reveal a structured landscape of black-hole masses and spins, suggesting multiple formation processes are now being resolved observationally. We analyze data from LIGO--Virgo--KAGRA's (LVK's) fifth Gravitational-Wave Transient Catalog and find that the population is naturally described by four distinct subpopulations. The dominant component, contributing $\simeq70\%$ of the astrophysical merger rate, is characterised by a low-mass population centred near $10M_\odot$ and is separated from heavier systems by a depletion near $14M_\odot$. This component may be associated with black holes formed from failed supernovae. Above this depletion, we find two intermediate-mass components: an unequal-mass branch pairing the lower-mass, $\simeq10M_\odot$ black hole with a heavier black hole, perhaps associated with isolated-binary/stable-mass-transfer formation, and a nearly equal-mass branch peaking near $30$--$35M_\odot$ whose low spins and mass distribution favour first-generation systems possibly born in dense stellar environments. A fourth, percent-level component extends to higher masses and is characterized by a broad mass-ratio distribution and large spin magnitudes, consistent with a hierarchical-merger population. Our four-component model is overwhelmingly preferred over a standard LVK population model by a natural-log Bayes factor of $\ln(\mathrm{BF}) = 19.2$. Our work observationally unveils a new subpopulation of black-hole mergers utilising a new hybrid data-driven and parametric-model discovery method, bringing us one step closer to understanding stellar-mass black-hole archaeology.

astro-ph.HE↗

arXiv2026

Joint inference for gravitational-wave signal and noise glitch: Method and application

Non-Gaussian noise transients ("glitches") in gravitational-wave observatories degrade our ability to accurately perform astrophysical inference. We present the analysis pipeline bilby_glitch, which allows for simultaneous Bayesian inference of gravitational-wave signals and glitches. Our framework is modular and built on top of the popular bilby framework, facilitating future extensions with additional glitch and signal models. We integrate transdimensional bilby into our framework and discuss three glitch models: a physically-motivated slow scattering model, and flexible sine-Gaussian and chirplet models. Using a combination of simulated and real data, we demonstrate that bilby_glitch produces reliable results. We then reanalyse two gravitational-wave events - GW191109 and GW200129 - which show signs of interesting black-hole spins, but which may also be affected by data-quality issues. Our results for GW191109 are consistent with previous analysis. For GW200129, we recover results consistent with Payne et al., where the evidence of spin-precession is much weaker when using the waveform approximant NRSur7dq4 in combination with wavelet-based glitch modeling. Furthermore, we show the astrophysical conclusion of this event is dependent on the interplay between the waveform approximant and glitch model, since in contrast to NRSur7dq4 we find that inference with the waveform approximant IMRPhenomXPHM shows strong evidence of spin-precession when used in combination with wavelet-based glitch modeling.

gr-qc↗

arXiv2026

Licence to Bin: Accurate and Scalable Inference for Binary Neutron Stars in Next-Generation Gravitational-Wave Detectors

Next-generation gravitational-wave observatories will observe binary neutron-star mergers with much higher signal-to-noise ratios, over much longer durations and across broader frequency bands than current detectors. These long-duration signals present a major computational challenge for Bayesian parameter estimation. Reduced-order quadrature is a promising approach for accelerating inference, but in this regime its standard construction encounters severe memory and accuracy limitations. We present a practical reduced-order quadrature construction for long binary neutron-star signals with time-dependent detector response and full effects of the observatories' free-spectral range. Our approach combines improved adaptive frequency sampling, disk-backed streaming, and subbanded reduced-order quadrature construction, enabling efficient and accurate reduced-order models for signals that were previously intractable. We demonstrate for the first time reduced-order Bayesian inference on an approximately 2 h binary neutron-star signal extending down to 5 Hz and with signal-to-noise ratio 2090. We show the resulting reduced-order quadrature remains sufficiently accurate for practical inference. The full analysis is carried out in about 48 h using 128 CPU cores. We also find that, when time-dependent detector-response effects are included, a single Cosmic Explorer detector can localize such a signal to a 90% credible sky area of approximately $41~\mathrm{deg}^2$, with important implications for multimessenger astronomy and cosmology. These results demonstrate that reduced-order methods can make next-generation binary neutron-star inference computationally feasible.

astro-ph.HE↗

arXiv2026

Fortifying gravitational-wave population inference with normalizing flows

As the LIGO-Virgo-KAGRA collaboration's (LVK's) gravitational-wave transient catalog grows, we are learning a wealth of information from the population properties of binary black hole mergers. Events in the catalog are represented with posterior samples describing the astrophysical parameters for each event. Population studies combine these samples to measure the distribution of astrophysical parameters such as black hole masses and spins. However, the posterior-sample representation of each event is only approximate. We construct a mock population with masses drawn from an astrophysically-motivated distribution with sharp features. Using this, we demonstrate that when $\gtrsim 300$ events are combined, even with each event's posterior represented by $1 \times 10^4 {-} 2 \times 10^4$ samples, the numerical error can become large enough that the resulting population inference is unreliable. We consider two solutions. In the short term, we show that nested samples (already produced by LVK analyses) can be used to more accurately describe each event in population studies. But this will only grant a temporary reprieve until the nested-sample representation becomes inadequate. In the longer term, we propose to represent each event with a normalizing flow. In order to represent each event with sufficient accuracy, each normalizing flow can be used to generate an arbitrarily large number of new posterior samples with a significantly reduced computational cost relative to traditional sampling methods. When compared to nested sampling, our normalizing flows produce posterior draws with a median of $\approx 80\%$ fewer likelihood evaluations per sample, while also providing greater opportunity for parallelization. We believe refinement of normalizing flow architectures and training techniques in future works could further reduce this per-sample cost significantly.

astro-ph.HE↗

arXiv2026

A universal framework to identify eccentric binary mergers: GW200105 case study

Orbital eccentricity in gravitational-wave signals from merging compact object binaries is a powerful indicator of their formation channel. Several binary black hole mergers and a neutron star--black hole merger have been reported to exhibit signs of eccentricity, but which events are identified and the significance of the eccentricity differs between studies. Measurements of eccentricity can change depending on the choice of prior. The choice of prior is subtle: eccentricity is commonly measured at an arbitrary reference frequency, which varies from study to study. We use the candidate eccentric neutron star--black hole merger GW200105_162426 as a case study, employing a range of priors and reference frequencies, and find the results to be strongly prior-driven. We show that the varied results reported across different studies can be partially reconciled by accounting for the evolution of eccentricity with reference frequency. In order to make conclusive statements about eccentricity, we propose a detection statistic that does not depend on reference frequency, and which marginalises over astrophysically-motivated distributions in eccentricity. Using this detection statistic, we find reduced support for the eccentric hypothesis for GW200105_162426: we obtain a natural log Bayes factor ln B $\leq$ 0.9 comparing the eccentric, aligned-spin hypothesis to the quasi-circular, precessing hypothesis. Our results cast doubt on the eccentric interpretation of GW200105_162426 and underscore the importance of modelling the astrophysical distributions of eccentricity in nature.

astro-ph.HE↗

arXiv2026

A comprehensive framework for phase-coherent mapping of the gravitational-wave sky with pulsar timing arrays

We present a practical implementation of a phase-coherent mapping technique for pulsar timing arrays that resolves the full complex polarisation state of the gravitational-wave sky as a function of direction and frequency. Unlike standard cross-correlation methods, this approach preserves the amplitude, phase, and polarisation of the signal in every sky pixel. The resulting maps constitute a compact, minimally processed summary of the data from which all subsequent analyses -- characterisation of a stochastic background, searches for anisotropy, and identification of individual sources -- can be derived within a single unified framework. Our implementation is fully compatible with established pulsar timing data analysis methods. We validate the framework through a series of realistic simulations with varying array configurations, noise properties, and signal types. We demonstrate robust recovery of source amplitudes and sky locations across different scenarios, and discuss the impact of polarisation leakage, noise, and direction-dependent array sensitivity on the recovery of astrophysical signals.

astro-ph.HE↗

arXiv2026

Enhancing Event Reconstruction in Hyper-Kamiokande with Machine Learning: A ResNet Implementation

The forthcoming Hyper-Kamiokande experiment requires substantially larger Monte Carlo datasets than previous experiments to satisfy stringent systematic-uncertainty requirements. While traditional maximum-likelihood reconstruction provides high-quality results, its per-event computational cost makes processing these large samples increasingly impractical. We demonstrate a neural-network-based reconstruction approach for the Hyper-Kamiokande far detector using simulated data. Single-particle events with kinetic energies from the Cherenkov threshold up to 2 GeV are propagated through the detector, with PMT charge and timing information mapped to $190\times189$ two-channel images serving as inputs to ResNet models in the WatChMaL framework. These models (i) classify events into four particle hypotheses ($e$, $\mu$, $\gamma$, $\pi^{0}$) and (ii) regress the vertex, direction, and momentum of electrons and muons. Averaged over the full kinematic range, the regression models achieve momentum resolutions of $1.35\%$ and $2.39\%$, angular resolutions of $1.25^\circ$ and $1.94^\circ$, and vertex resolutions of $28.2$ cm and $25.4$ cm, for muons and electrons respectively, broadly consistent with traditional methods. The classifier improves $e$-$\mu$, $e$-$\gamma$, and $e$-$\pi^{0}$ separation, with ROC curve areas of $0.9999992$, $0.633$, and $0.9526$. Crucially, our networks achieve inference times of 1-2 ms per event on a single GPU, yielding speed-ups of $3.2\times10^{4}$-$5.2\times10^{4}$ relative to likelihood-based reconstruction, highlighting deep learning as a scalable alternative for Hyper-Kamiokande event reconstruction.

hep-ex↗

arXiv2026

A transdimensional sampling framework for pulsar timing noise modelling

A careful characterisation of the noise processes in pulsar timing data is a prerequisite for pulsar timing array experiments. While single-pulsar noise analyses are crucial for both gravitational-wave searches and astrophysical studies, they are often computationally intensive and rely on running and comparing multiple fixed noise models. We present tPTABilby, a transdimensional Bayesian inference framework for single-pulsar noise analysis built on the Bilby library. The method flexibly models a wide range of noise processes like radiometer noise, pulse-phase jitter, intrinsic red noise, dispersion measure variations, and chromatic interstellar medium effects. By employing transdimensional sampling, tPTABilby simultaneously infers the number and type of active noise sources, providing a unified treatment of model selection and parameter estimation. We validate the methodology through simulations with known injected noise models, demonstrating accurate recovery of model probabilities and calibrated posterior distributions. We then apply this approach to a single pulsar, PSR J1713+0747, from a MeerKAT Pulsar Timing Array (MPTA) dataset, analysing the data with both tPTABilby and Enterprise, and subsequently compare the results with existing MPTA analyses through posterior predictive checks of the inferred noise spectra. Our results highlight the flexibility of transdimensional approaches to single-pulsar noise analysis, demonstrating consistency with standard fixed-model methods while providing a more statistically robust framework, and present tPTABilby as a simple and reproducible approach for PTA inference.

astro-ph.IM↗

arXiv2026

Searching for gravitational waves from compact binary mergers powering long gamma-ray bursts during LIGO-Virgo-KAGRA's O3 run

Neutron star binary mergers are often associated with short gamma-ray bursts (GRBs), but the recent detection of kilonovae coincident with long GRBs suggest that some mergers may produce long GRBs. Motivated by these developments, we perform a search for binary neutron star and neutron star-black hole gravitational-wave signals coincident with long GRBs using data from the third LIGO--Virgo--KAGRA (LVK) observing run. We analyze LVK data coincident with long GRBs detected by Fermi's GRB Monitor and Swift's Burst Alert Telescope when at least two gravitational-wave observatories were running. We find no evidence of a coincident gravitational-wave signal and set limits on the luminosity distance to each of these long GRBs under the assumption that they were powered by binary mergers.

astro-ph.HE↗

arXiv2026

Optimising gravitational-wave sky maps for pulsar timing arrays

Pulsar timing arrays (PTAs) have recently reported compelling evidence for the presence of a gravitational-wave background signal. Mapping the gravitational-wave background is key to understanding how it is formed, since anisotropy is a tracer for, for example, a supermassive black hole binary origin. In this work we refine the frequentist regularised gravitational-wave mapping analysis developed in our previous work (as part of the MeerKAT PTA 4.5-year data release). We derive a point-spread function describing the angular resolution of a PTA. We investigate how the point spread function changes for different PTA constellations and determine the best possible angular resolution achievable within our framework. Using simulated data, we demonstrate that previous methods do not capture the actual resolution - especially in regions of the sky with a high density of pulsars. We propose an improved scheme that accounts for a variable local resolution and test it using realistic simulations of the latest MeerKAT dataset. We demonstrate that we are able to identify a continuous gravitational wave signal in a region with good pulsar sky coverage with approximately a factor of two increase in significance compared to our previous method.

astro-ph.IM↗

arXiv2025

Frequency- and phase-resolved polarimetry of millisecond pulsars and its application to timing

Pulsar timing is used for a variety of applications including tests of fundamental physics, probing the structure of neutron stars, and detecting nanohertz gravitational waves. Development of robust methods and generation of high-quality timing data is therefore of utmost importance. In this paper, we present a new technique for creating high-fidelity templates that can be used to measure the pulse times of arrival with significantly increased precision compared to existing methods. Our framework makes use of all available polarimetric information to generate frequency-dependent models of pulse-shape evolution of all four Stokes parameters. We apply this method to millisecond pulsars observed by the Parkes Pulsar Timing Array and show that it results in timing measurement uncertainties reduced up to $\sim$20-30%. We also present, for the first time, phase- and frequency-resolved polarimetric measurements of millisecond pulsars observed with the Parkes Murriyang ultra-widebandwith-low receiver. The data, plots and the code underlying this analysis are made publicly available.

astro-ph.HE↗

arXiv2025

A subpopulation of low-mass, spinning black holes: signatures of dynamical assembly

Gravitational-wave observations of massive, rapidly spinning binary black holes mergers provide increasing evidence for the dynamical origin of some mergers. Previous studies have interpreted the mergers with primary mass $\gtrsim45\,M_\odot$ as being dominated by hierarchical, second-generation mergers, with rapidly spinning primaries being the products of previous black hole mergers assembled in dense stellar clusters. In this work, we reveal confident evidence of another subpopulation with rapid and isotropic spins at low mass containing the two exceptional events GW241011 and GW241110, consistent with a hierarchical merger hypothesis. Our result suggests the mass distribution of the second-generation black holes is peaked at low primary masses of $\sim16\,M_\odot$ rather than $\gtrsim45\,M_\odot$ in the pair-instability gap. Such low-mass second-generation black holes must be formed from the merger of even lighter first-generation black holes, implying that dense, metal-rich stellar environments contribute to the binary black hole population. By separating the contamination of higher-generation black holes, our result reveals the primary mass distribution of first-generation black holes formed from stellar collapse, which shows a significant dip between $\sim12\,M_\odot$ to $\sim20\,M_\odot$. This may indicate a dearth of black holes due to variation in the core compactness of the progenitor.

astro-ph.HE↗

arXiv2025

Is GW231123 a hierarchical merger?

The binary black hole merger GW231123 is both the most massive gravitational-wave event observed and has the highest component spins measured to date. The dimensionless spins of the more massive (primary) and less massive (secondary) black holes are measured to be $\chi_1 = 0.90^{+0.10}_{-0.19}$ and $\chi_2 = 0.80^{+0.20}_{-0.51}$ ($90\%$ credible intervals), respectively. Its large mass and extremal spins are challenging to explain through standard binary stellar physics, though a flurry of hypothetical scenarios have been proposed. Hierarchical assembly - i.e., mergers of black holes that are themselves formed from previous generations of mergers - is generally a promising way to explain massive and rapidly spinning black holes. Here, we investigate the possibility that GW231123 was assembled hierarchically in a dense star cluster as the merger of two second-generation black holes. Taking the inferred spin values at face value, we find that it is possible ($p\approx 5\%$) that a compact binary with component spins like GW231123 could form in a cluster from hierarchical assembly.

astro-ph.HE↗

arXiv2025

GW231123: extreme spins or microglitches?

The recently reported binary black hole merger, GW231123, has unusual properties that make it hard to explain astrophysically. Parameter estimation studies are consistent with maximally spinning black holes and the dimensionless spin of the more massive component is constrained to be $\chi_1\gtrsim 0.8$. Analysis of data also revealed potential systematics that could not be fully replicated with simulated studies. We explore the possibility that these measurements are biased due to unmodeled non-Gaussian noise in the detectors, and that the actual black hole spins are more modest. We present evidence for a population of \textit{microglitches} in LIGO gravitational-wave strain data that can lead to biases in the parameter estimation of short-duration signals such as GW231123. Using simulated data of a massive event like GW231123, we demonstrate how microglitches can bias our measurements of black hole spins toward $\chi\approx1$ with negligible posterior support for the true value of $\chi\approx0.7$. We develop a noise model to account for microglitches and show that this model successfully reduces biases in the recovery of signal parameters. We characterize the microglitch population in real interferometer data surrounding GW231123 and find a single detector glitch duty cycle of $0.57_{-0.19}^{+0.21}$, which implies nearly a $100\%$ probability that at least one event through the fourth gravitational wave transient catalog coincides with microglitches in two detectors. We argue that further investigations are required before we can have a confident picture of the astrophysical properties of GW231123.

gr-qc↗

arXiv2025

Evidence for Three Subpopulations of Merging Binary Black Holes at Different Primary Masses

With the release of the fourth LIGO--Virgo--KAGRA gravitational-wave catalog (GWTC-4), we are starting to gain a detailed view of the population of merging binary black holes. The formation channels of these black holes is not clearly understood, but different formation mechanisms may lead to subpopulations with different properties visible in gravitational-wave data. Adopting a phenomenological approach, we find GWTC-4 data supports the presence of at least three subpopulations, each associated with a different range of black hole mass and with sharp transition boundaries between them. Each subpopulation is characterized by different distributions for either the mass ratios, the black-hole spin magnitudes or both. Subpopulation A with primary mass $m_1 \leq 27.7^{+4.1}_{-3.4} M_{\odot}$ ($90 \%$ credibility), is characterized by a nearly flat mass ratio distribution $q=m_2/m_1$, and by small spin magnitudes ($\chi \leq 0.5^{+0.1}_{-0.1}$). Subpopulation B with $27.7^{+4.1}_{-3.4} M_{\odot} \leq m_1 \leq 40.2^{+4.7}_{-3.2} M_{\odot}$, has a much sharper preference for mass ratio $q \approx 1$. Subpopulation C, with $m_1 \geq 40.2^{+4.7}_{-3.2} M_{\odot}$, has support for large spin magnitudes, and tentative support for mass ratios $q\approx0.5$. We interpret these transitions as evidence for multiple subpopulations, each potentially associated with a different formation pathways. We suggest potential formation scenarios for each subpopulations, and suggest that Subpopulation B may be associated with chemically homogeneous evolution or population III stars. Our findings for Subpopulation C are largely consistent with recent claims of hierarchical mergers, but with some curious differences in properties.

astro-ph.HE↗

arXiv2025

Trends in the Population of Binary Black Holes Following the Fourth Gravitational-Wave Transient Catalog: a Data-Driven Analysis

Current population models of binary black hole distributions are difficult to interpret because standard population inferences hinge on modeling choices, which can mask or mimic real structure. The maximum population likelihood ``$\pistroke$ formalism'' provides a means to investigate and interpret features in the distribution of binary black holes using only data -- without specifying a population model. It tells us if features inferred from current population models are truly present in the data or if they arise from model misspecification. It also provides guidance for developing new models by highlighting previously unnoticed features. In this study, we utilize the $\pistroke$ formalism to examine the binary black hole population in the LIGO--Virgo--KAGRA (LVK) fourth Gravitational-Wave Transient Catalog (GWTC-4). Our analysis supports the existence of a gap around $45\,M_\odot$ in the secondary black hole mass distribution and identifies a widening in the distribution of the effective inspiral spin parameter $\chi_\text{eff}$ near this mass as recently reported by Tong et al. (2025). Similar to earlier studies, we find support for an anti-correlation between $\chi_\text{eff}$ and mass ratio. However, we argue that this may be a spurious correlation arising from misspecification of the joint distribution of black hole masses. Furthermore, we identify support for dimensionless black hole spin magnitudes at approximately $\chi \approx 0.2$ and $\chi\approx0.7$. The data support the existence of a correlation between the spin magnitudes $\chi_1$ and $\chi_2$, though subsequent study is required to determine if this feature is statistically significant. The accompanying data release includes $\pistroke$ samples, which can be used to compare theoretical predictions to LVK data and to assess assumptions in parameterised models.

astro-ph.HE↗