Hidden Markov model and cross-correlation searches for continuous gravitational waves

Abstract

Gravitational waves, perturbations in space-time propagating at the speed of light, provide information about astrophysical sources. In recent years, transient gravitational-wave signals from compact binary coalescences have been observed directly by two ground-based detectors, the Advanced Laser Interferometer Gravitational Wave Observatory (Advanced LIGO) and the Virgo detector. In addition to the transient signals, searches are conducted for weaker, persistent, quasimonochromatic signals from rapidly spinning neutron stars. In this thesis, we mainly focus on searches for two interesting categories of persistent gravitational-wave sources, low-mass X-ray binaries and young supernova remnants. We describe two data analysis methods, based on hidden Markov model (HMM) and cross-correlation algorithms, conduct Monte-Carlo simulations to validate the methods, and present search results using both Initial LIGO and Advanced LIGO data. No evidence of continuous gravitational waves is found, but the analyses in this thesis improve substantially on previously published upper limits.

The HMM scheme provides a strategy for tracking wandering of the unknown neutron star spin frequency in continuous-wave searches. It was applied to search for continuous waves from the brightest low-mass X-ray binary, Scorpius X-1, analyzing the data collected during the first Advanced LIGO observing run (O1). It yields frequentist 95% confidence strain upper limits h_0^95% = 4.0e-25, 8.3e-25, and 3.0e-25 for electromagnetically restricted source orientation, unknown polarization, and circular polarization, respectively, at 106 Hz. They are <= 10 times larger than the theoretical torque-balance strain limit. A sideband search, previously applied to Initial LIGO Science Run 5, was conducted for the same source using one 10-day data stretch in O1. Bayesian analytic median strain upper limits h_0^95% = 3.5e-25 and h_0^95% = 2.2e-25 are reported for the unknown polarization and electromagnetically restricted source orientation, respectively, in the 100-200 Hz band. These analytic upper limits are less conservative than the empirical ones, but improve on S5 upper limits by a factor of 3.6.

A cross-correlation search is also presented for the young supernova remnant, SNR 1987A, using the second year of Initial LIGO Science Run 5 data. It yields a 90% confidence upper limit h_0^90% = 3.8e-25, corresponding to an ellipticity of epsilon = 8.2e-4, near 150 Hz. The cross-correlation search is computationally demanding, creating a need for more economical alternatives. One such algorithm, based on the HMM tracking scheme, is developed and validated. It tracks the rapid phase evolution of young objects from secular braking and stochastic timing noise torques simultaneously without searching second- and higher-order derivatives of the signal frequency. It is expected to be applied to upcoming searches for young objects, including post-merger remnants of binary neutron star mergers, like GW170817.