TDI-infinity: time-delay interferometry without delays

The space-based gravitational-wave observatory LISA relies on a form of synthetic interferometry (time-delay interferometry, or TDI) where the otherwise overwhelming laser phase noise is canceled by linear combinations of appropriately delayed phase measurements. These observables grow in length and complexity as the realistic features of the LISA orbits are taken into account. In this paper we outline an implicit formulation of TDI where we write the LISA likelihood directly in terms of the basic phase measurements, and we marginalize over the laser phase noises in the limit of infinite laser-noise variance. Equivalently, we rely on TDI observables that are defined numerically (rather than algebraically) from a discrete-filter representation of the laser propagation delays. Our method generalizes to any time dependence of the armlengths; it simplifies the modeling of gravitational-wave signals; and it allows a straightforward treatment of data gaps and missing measurements.

On parametric tests of relativity with false degrees of freedom

General relativity can be tested by comparing the binary-inspiral signals found in LIGO--Virgo data against waveform models that are augmented with artificial degrees of freedom. This approach suffers from a number of logical and practical pitfalls. 1) It is difficult to ascribe meaning to the stringency of the resultant constraints. 2) It is doubtful that the Bayesian model comparison of relativity against these artificial models can offer actual validation for the former. 3) It is unknown to what extent these tests might detect alternative theories of gravity for which there are no computed waveforms; conversely, when waveforms are available, tests that employ them will be superior.

Learning Bayesian posteriors with neural networks for gravitational-wave inference

We seek to achieve the Holy Grail of Bayesian inference for gravitational-wave astronomy: using deep-learning techniques to instantly produce the posterior $p(theta|D)$ for the source parameters $theta$, given the detector data $D$. To do so, we train a deep neural network to take as input a signal + noise data set (drawn from the astrophysical source-parameter prior and the sampling distribution of detector noise), and to output a parametrized approximation of the corresponding posterior. We rely on a compact representation of the data based on reduced-order modeling, which we generate efficiently using a separate neural-network waveform interpolant [A. J. K. Chua, C. R. Galley & M. Vallisneri, Phys. Rev. Lett. 122, 211101 (2019)]. Our scheme has broad relevance to gravitational-wave applications such as low-latency parameter estimation and characterizing the science returns of future experiments. Source code and trained networks are available online at https://github.com/vallis/truebayes.

Taming outliers in pulsar-timing datasets with hierarchical likelihoods and Hamiltonian sampling

Pulsar-timing datasets have been analyzed with great success using probabilistic treatments based on Gaussian distributions, with applications ranging from studies of neutron-star structure to tests of general relativity and searches for nanosecond gravitational waves. As for other applications of Gaussian distributions, outliers in timing measurements pose a significant challenge to statistical inference, since they can bias the estimation of timing and noise parameters, and affect reported parameter uncertainties. We describe and demonstrate a practical end-to-end approach to perform Bayesian inference of timing and noise parameters robustly in the presence of outliers, and to identify these probabilistically. The method is fully consistent (i.e., outlier-ness probabilities vary in tune with the posterior distributions of the timing and noise parameters), and it relies on the efficient sampling of the hierarchical form of the pulsar-timing likelihood. Such sampling has recently become possible with a no-U-turn Hamiltonian sampler coupled to a highly customized reparametrization of the likelihood; this code is described elsewhere, but it is already available online. We recommend our method as a standard step in the preparation of pulsar-timing-array datasets: even if statistical inference is not affected, follow-up studies of outlier candidates can reveal unseen problems in radio observations and timing measurements; furthermore, confidence in the results of gravitational-wave searches will only benefit from stringent statistical evidence that datasets are clean and outlier-free.

The LIGO Open Science Center

The LIGO Open Science Center (LOSC) fulfills LIGOs commitment to release, archive, and serve LIGO data in a broadly accessible way to the scientific community and to the public, and to provide the information and tools necessary to understand and use the data. In August 2014, the LOSC published the full dataset from Initial LIGOs S5 run at design sensitivity, the first such large-scale release and a valuable testbed to explore the use of LIGO data by non-LIGO researchers and by the public, and to help teach gravitational-wave data analysis to students across the world. In addition to serving the S5 data, the LOSC web portal (losc.ligo.org) now offers documentation, data-location and data-quality queries, tutorials and example code, and more. We review the mission and plans of the LOSC, focusing on the S5 data release.

Low-rank approximations for large stationary covariance matrices, as used in the Bayesian and generalized-least-squares analysis of pulsar-timing data

Many data-analysis problems involve large dense matrices that describe the covariance of stationary noise processes; the computational cost of inverting these matrices, or equivalently of solving linear systems that contain them, is often a practical limit for the analysis. We describe two general, practical, and accurate methods to approximate stationary covariance matrices as low-rank matrix products featuring carefully chosen spectral components. These methods can be used to greatly accelerate data-analysis methods in many contexts, such as the Bayesian and generalized-least-squares analysis of pulsar-timing residuals.

New advances in the Gaussian-process approach to pulsar-timing data analysis

In this work we review the application of the theory of Gaussian processes to the modeling of noise in pulsar-timing data analysis, and we derive various useful and optimized representations for the likelihood expressions that are needed in Bayesian inference on pulsar-timing-array datasets. The resulting viewpoint and formalism lead us to two improved parameter-sampling schemes inspired by Gibbs sampling. The new schemes have vastly lower chain autocorrelation lengths than the Markov Chain Monte Carlo methods currently used in pulsar-timing data analysis, potentially speeding up Bayesian inference by orders of magnitude. The new schemes can be used for a full-noise-model analysis of the large datasets assembled by the International Pulsar Timing Array collaboration, which present a serious computational challenge to existing methods.

Bayesian inference for pulsar timing models

The extremely regular, periodic radio emission from millisecond pulsars makes them useful tools for studying neutron star astrophysics, general relativity, and low-frequency gravitational waves. These studies require that the observed pulse times of arrival be fit to complex timing models that describe numerous effects such as the astrometry of the source, the evolution of the pulsars spin, the presence of a binary companion, and the propagation of the pulses through the interstellar medium. In this paper, we discuss the benefits of using Bayesian inference to obtain pulsar timing solutions. These benefits include the validation of linearized least-squares model fits when they are correct, and the proper characterization of parameter uncertainties when they are not; the incorporation of prior parameter information and of models of correlated noise; and the Bayesian comparison of alternative timing models. We describe our computational setup, which combines the timing models of Tempo2 with the nested-sampling integrator MultiNest. We compare the timing solutions generated using Bayesian inference and linearized least-squares for three pulsars: B1953+29, J2317+1439, and J1640+2224, which demonstrate a variety of the benefits that we posit.

The Gravitational-Wave Discovery Space of Pulsar Timing Arrays

Recent years have seen a burgeoning interest in using pulsar timing arrays (PTAs) as gravitational-wave (GW) detectors. To date, that interest has focused mainly on three particularly promising source types: supermassive--black-hole binaries, cosmic strings, and the stochastic background from early-Universe phase transitions. In this paper, by contrast, our aim is to investigate the PTA potential for discovering unanticipated sources. We derive significant constraints on the available discovery space based solely on energetic and statistical considerations: we show that a PTA detection of GWs at frequencies above ~3.e-5 Hz would either be an extraordinary coincidence or violate cherished beliefs; we show that for PTAs GW memory can be more detectable than direct GWs, and that, as we consider events at ever higher redshift, the memory effect increasingly dominates an events total signal-to-noise ratio. The paper includes also a simple analysis of the effects of pulsar red noise in PTA searches, and a demonstration that the effects of periodic GWs in the 10^-8 -- 10^-4.5 Hz band would not be degenerate with small errors in standard pulsar parameters (except in a few narrow bands).

Stealth Bias in Gravitational-Wave Parameter Estimation

Inspiraling binaries of compact objects are primary targets for current and future gravitational-wave observatories. Waveforms computed in General Relativity are used to search for these sources, and will probably be used to extract source parameters from detected signals. However, if a different theory of gravity happens to be correct in the strong-field regime, source-parameter estimation may be affected by a fundamental bias: that is, by systematic errors induced due to the use of waveforms derived in the incorrect theory. If the deviations from General Relativity are not large enough to be detectable on their own and yet these systematic errors remain significant (i.e., larger than the statistical uncertainties in parameter estimation), fundamental bias cannot be corrected in a single observation, and becomes stealth bias. In this article we develop a scheme to determine in which cases stealth bias could be present in gravitational-wave astronomy. For a given observation, the answer depends on the detection signal-to-noise ratio and on the strength of the modified-gravity correction. As an example, we study three representative stellar-mass binary systems that will be detectable with second-generation ground-based observatories. We find that significant systematic bias can occur whether or not modified gravity can be positively detected, for correction strengths that are not currently excluded by any other experiment. Thus, stealth bias may be a generic feature of gravitational-wave detections, and it should be considered and characterized, using expanded models such as the parametrized post-Einstein framework, when interpreting the results of parameter-estimation analyses.