-
Iteratively Comparing Gravitational-Wave Observations to the Evolution of Massive Stellar Binaries
Authors:
Vera Delfavero,
Richard O'Shaughnessy,
Krzysztof Belczynski,
Paweł Drozda,
Daniel Wysocki
Abstract:
Gravitational-wave observations have the capability to strongly differentiate between different assumptions for how binary compact objects form. In this work, we show how to carefully interpolate a marginal likelihood between choices of binary evolution model parameters. Using the StarTrack binary evolution code, we compare one- and four-dimensional binary evolution models to the compact binary me…
▽ More
Gravitational-wave observations have the capability to strongly differentiate between different assumptions for how binary compact objects form. In this work, we show how to carefully interpolate a marginal likelihood between choices of binary evolution model parameters. Using the StarTrack binary evolution code, we compare one- and four-dimensional binary evolution models to the compact binary mergers reported in recent gravitational-wave observing runs. We first consider a one-dimensional model, studying the effect of supernova kick velocity (drawn from a Maxwellian with dispersion σ_eff) on the simulated population of compact binary mergers, and find support for substantial SN recoil kicks. We follow this up with a four-dimensional study of σ_eff, mass transfer efficiency (fa) and the efficiency of angular momentum depletion from ejected material (\b{eta}) during Roche-lobe accretion, and an observation-driven reduction in the mass-loss rate estimated from stellar wind models (fwind1). We find that three of them (σ_eff, fa, and fwind1) can be efficiently limited by these observational comparisons. After initially sampling from a uniform prior in the space of these parameters, we refined our sampling by iteratively estimating a Bayesian likelihood for each simulation and fitting that likelihood to a parametric model (a truncated Gaussian) in order to propose new. Our maximum likelihood simulation (K0559) has parameters: σ_eff = 108.3 km/s (indicating substantial SN recoil kicks), fa = 0.922 (indicating efficient mass transfer), and fwind1 = 0.328 (indicating support for reduced wind-driven mass loss). Note that our estimates are only valid within one particular model of compact binary formation through isolated binary evolution and do not yet take into account the impact of other uncertain pieces of stellar physics and binary evolution.
△ Less
Submitted 7 July, 2023; v1 submitted 9 March, 2023;
originally announced March 2023.
-
Low-latency parameter inference enabled by a Gaussian likelihood approximation for RIFT
Authors:
A. B. Yelikar,
V. Delfavero,
R. O'Shaughnessy
Abstract:
Rapid identification, characterization, and localization of gravitational waves from binary compact object mergers can enable well-informed follow-on multimessenger observations. In this work, we investigate a small modification to the RIFT parameter inference pipeline to enable extremely low-latency inference, tested here for nonprecessing sources.
Rapid identification, characterization, and localization of gravitational waves from binary compact object mergers can enable well-informed follow-on multimessenger observations. In this work, we investigate a small modification to the RIFT parameter inference pipeline to enable extremely low-latency inference, tested here for nonprecessing sources.
△ Less
Submitted 3 January, 2023;
originally announced January 2023.
-
Expanding RIFT: Improving performance for GW parameter inference
Authors:
J. Wofford,
A. Yelikar,
H. Gallagher,
E. Champion,
D. Wysocki,
V. Delfavero,
J. Lange,
C. Rose,
V. Valsan,
S. Morisaki,
J. Read,
C. Henshaw,
R. O'Shaughnessy
Abstract:
The Rapid Iterative FiTting (RIFT) parameter inference algorithm provides a framework for efficient, highly-parallelized parameter inference for GW sources. In this paper, we summarize essential algorithm enhancements and operating point choices for the RIFT iterative algorithm, including choices used for analysis of LIGO/Virgo O3 observations. We also describe other extensions to the RIFT algorit…
▽ More
The Rapid Iterative FiTting (RIFT) parameter inference algorithm provides a framework for efficient, highly-parallelized parameter inference for GW sources. In this paper, we summarize essential algorithm enhancements and operating point choices for the RIFT iterative algorithm, including choices used for analysis of LIGO/Virgo O3 observations. We also describe other extensions to the RIFT algorithm and software ecosystem. Some extensions increase RIFT's flexibility to produce outputs pertinent to GW astrophysics. Other extensions increase its computational efficiency or stability. Using many randomly-selected sources, we assess code robustness with two distinct code configurations, one designed to mimic settings as of LIGO O3 and another employing several performance enhancements. We illustrate RIFT's capabilities with analysis of selected events.
△ Less
Submitted 1 February, 2023; v1 submitted 14 October, 2022;
originally announced October 2022.
-
Compressed Parametric and Non-Parametric Approximations to the Gravitational Wave Likelihood
Authors:
Vera Delfavero,
Richard O'Shaughnessy,
Daniel Wysocki,
Anjali Yelikar
Abstract:
Gravitational-wave observations of quasicircular compact binary mergers imply complicated posterior measurements of their parameters. Though Gaussian approximations to the pertinent likelihoods have decades of history in the field, the relative generality and practical utility of these approximations hasn't been appreciated, given focus on careful, comprehensive generic Bayesian parameter inferenc…
▽ More
Gravitational-wave observations of quasicircular compact binary mergers imply complicated posterior measurements of their parameters. Though Gaussian approximations to the pertinent likelihoods have decades of history in the field, the relative generality and practical utility of these approximations hasn't been appreciated, given focus on careful, comprehensive generic Bayesian parameter inference. Building on our previous work in three dimensions, we demonstrate by example that bounded multivariate normal likelihood approximations are a sufficiently accurate representation of the full likelihood of observed gravitational-wave sources. Fits for each event published in the Gravitatinoal-Wave Transient Catalogs at https://gitlab.com/xevra/nal-data, along with a code release at https://gitlab.com/xevra/gwalk. We argue our approximations are more than accurate enough for popultion inference and introduce much smaller errors than waveform model systematics. To demonstrate the utility of these approximations as parametric models for the likelihood of individual gravitational-wave sources, we show examples of their application to modeling the population of observed gravitational-wave sources as well as low-latency parameter inference.
△ Less
Submitted 6 December, 2022; v1 submitted 27 May, 2022;
originally announced May 2022.
-
A Nuclear Equation of State Inferred from Stellar r-process Abundances
Authors:
Erika M. Holmbeck,
Richard O'Shaughnessy,
Vera Delfavero,
Krzysztof Belczynski
Abstract:
Binary neutron star mergers (NSMs) have been confirmed as one source of the heaviest observable elements made by the rapid neutron-capture (r-) process. However, modeling NSM outflows -- from the total ejecta masses to their elemental yields -- depends on the unknown nuclear equation of state (EOS) that governs neutron-star structure. In this work, we derive a phenomenological EOS by assuming that…
▽ More
Binary neutron star mergers (NSMs) have been confirmed as one source of the heaviest observable elements made by the rapid neutron-capture (r-) process. However, modeling NSM outflows -- from the total ejecta masses to their elemental yields -- depends on the unknown nuclear equation of state (EOS) that governs neutron-star structure. In this work, we derive a phenomenological EOS by assuming that NSMs are the dominant sources of the heavy element material in metal-poor stars with r-process abundance patterns. We start with a population synthesis model to obtain a population of merging neutron star binaries and calculate their EOS-dependent elemental yields. Under the assumption that these mergers were responsible for the majority of r-process elements in the metal-poor stars, we find parameters representing the EOS for which the theoretical NSM yields reproduce the derived abundances from observations of metal-poor stars. For our proof-of-concept assumptions, we find an EOS that is slightly softer than, but still in agreement with, current constraints, e.g., by the Neutron Star Interior Composition Explorer, with $R_{1.4}=12.25\pm 0.03$~km and $M_{\textrm TOV}$ of $2.17\pm 0.03$~M$_\odot$(statistical uncertainties, neglecting modeling systematics).
△ Less
Submitted 28 February, 2022; v1 submitted 12 October, 2021;
originally announced October 2021.
-
Normal Approximate Likelihoods to Gravitational Wave Events
Authors:
Vera Delfavero,
Richard O'Shaughnessy,
Daniel Wysocki,
Anjali Yelikar
Abstract:
Gravitational wave observations of quasicircular compact binary mergers in principle provide an arbitrarily complex likelihood over eight independent intrinsic parameters: the masses and spins of the two merging objects. In this work, we demonstrate by example that a simple normal approximation over fewer (usually, three) effective dimensions provides a very accurate representation of the likeliho…
▽ More
Gravitational wave observations of quasicircular compact binary mergers in principle provide an arbitrarily complex likelihood over eight independent intrinsic parameters: the masses and spins of the two merging objects. In this work, we demonstrate by example that a simple normal approximation over fewer (usually, three) effective dimensions provides a very accurate representation of the likelihood, and allows us to replicate the eight-dimensional posterior over the mass and spin degrees of freedom. Alongside this paper, we provide the parameters for multivariate normal fits for each event published in GWTC-1 and GWTC-2, using the posterior samples from the catalog for each associated release. These normal approximations provide a highly efficient way to characterize gravitational wave observations when combining large numbers of events.
△ Less
Submitted 3 January, 2022; v1 submitted 27 July, 2021;
originally announced July 2021.
