GW190521: A dynamical capture of two black holes

We analyze the gravitational-wave signal GW190521 under the hypothesis that it was generated by the merger of two nonspinning black holes on hyperbolic orbits. The best configuration matching the data corresponds to two black holes of source frame masses of $81^{+62}_{-25}M_odot$ and $52^{+32}_{-32}M_odot$ undergoing two encounters and then merging into an intermediate-mass black hole. Under the hyperbolic merger hypothesis, we find an increase of one unit in the recovered signal-to-noise ratio and a 14 e-fold increase in the maximum likelihood value compared to a quasi-circular merger with precessing spins. We conclude that our results support the first gravitational-wave detection from the dynamical capture of two stellar-mass black holes.

${tt bajes}$: Bayesian inference of multimessenger astrophysical data, methods and application to gravitational-waves

We present ${tt bajes}$, a parallel and lightweight framework for Bayesian inference of multimessenger transients. ${tt bajes}$ is a Python modular package with minimal dependencies on external libraries adaptable to the majority of the Bayesian models and to various sampling methods. We describe the general workflow and the parameter estimation pipeline for compact-binary-coalescence gravitational-wave transients. The latter is validated against injections of binary black hole and binary neutron star waveforms, including confidence interval tests that demonstrates the inference is well-calibrated. Binary neutron star postmerger injections are also studied using a network of five detectors made of LIGO, Virgo, KAGRA and Einstein Telescope. Postmerger signals will be detectable for sources at ${lesssim}80,$Mpc, with Einstein Telescope contributing over 90% of the total signal-to-noise ratio. As a full scale application, we re-analyze the GWTC-1 black hole transients using the effective-one-body ${tt TEOBResumS}$ approximant, and reproduce selected results with other approximants. ${tt bajes}$ inferences are consistent with previous results; the direct comparison of ${tt bajes}$ and ${tt bilby}$ analyses of GW150914 shows a maximum Jensen-Shannon divergence of $5.2{times}10^{-4}$. GW170817 is re-analyzed using ${tt TaylorF2}$ with 5.5PN point-mass and 7.5PN tides, ${tt TEOBResumSPA}$, and ${tt IMRPhenomPv2_NRTidal}$ with different cutoff-frequencies of $1024,$Hz and $2048,$Hz. We find that the former choice minimizes systematics on the reduced tidal parameter, while a larger amount of tidal information is gained with the latter choice. ${tt bajes}$ can perform these analyses in about 1~day using 128 CPUs.

Machine Learning Gravitational Waves from Binary Black Hole Mergers

We apply machine learning methods to build a time-domain model for gravitational waveforms from binary black hole mergers, called mlgw. The dimensionality of the problem is handled by representing the waveforms amplitude and phase using a principal component analysis. We train mlgw on about $mathcal{O}(10^3)$ TEOBResumS and SEOBNRv4 effective-one-body waveforms with mass ratios $qin[1,20]$ and aligned dimensionless spins $sin[-0.80,0.95]$. The resulting models are faithful to the training sets at the ${sim}10^{-3}$ level (averaged on the parameter space). The speed up for a single waveform generation is a factor 10 to 50 (depending on the binary mass and initial frequency) for TEOBResumS and approximately an order of magnitude more for SEOBNRv4. Furthermore, mlgw provides a closed form expression for the waveform and its gradient with respect to the orbital parameters; such an information might be useful for future improvements in GW data analysis. As demonstration of the capabilities of mlgw to perform a full parameter estimation, we re-analyze the public data from the first GW transient catalog (GWTC-1). We find broadly consistent results with previous analyses at a fraction of the cost, although the analysis with spin aligned waveforms gives systematic larger values of the effective spins with respect to previous analyses with precessing waveforms. Since the generation time does not depend on the length of the signal, our model is particularly suitable for the analysis of the long signals that are expected to be detected by third-generation detectors. Future applications include the analysis of waveform systematics and model selection in parameter estimation.