Bayesian analysis on gravitational waves and exoplanets

Open Access
Author:
Deng, Xihao
Graduate Program:
Physics
Degree:
Doctor of Philosophy
Document Type:
Dissertation
Date of Defense:
December 01, 2014
Committee Members:
  • Lee S Finn, Dissertation Advisor
  • Lee S Finn, Committee Chair
  • Alexander Wolszczan, Committee Member
  • Stephane Coutu, Committee Member
  • Chad Richard Hanna, Committee Member
Keywords:
  • Bayesian analysis
  • gravitational waves
  • exoplanets
Abstract:
Attempts to detect gravitational waves using a pulsar timing array (PTA), i.e., a collection of pulsars in our Galaxy, have become more organized over the last several years. PTAs act to detect gravitational waves generated from very distant sources by observing the small and correlated effect the waves have on pulse arrival times at the Earth. In this thesis, I present advanced Bayesian analysis methods that can be used to search for gravitational waves in pulsar timing data. These methods were also applied to analyze a set of radial velocity (RV) data collected by the Hobby-Eberly Telescope on observing a K0 giant star. They confirmed the presence of two Jupiter mass planets around a K0 giant star and also characterized the stellar p-mode oscillation. The first part of the thesis investigates the effect of wavefront curvature on a pulsar's response to a gravitational wave. In it we show that we can assume the gravitational wave phasefront is planar across the array only if the source luminosity distance $\gg 2\pi L^2/\lambda$, where $L$ is the pulsar distance to the Earth ($\sim$ kpc) and $\lambda$ is the radiation wavelength ($\sim$ pc) in the PTA waveband. Correspondingly, for a point gravitational wave source closer than $\sim 100$ Mpc, we should take into account the effect of wavefront curvature across the pulsar-Earth line of sight, which depends on the luminosity distance to the source, when evaluating the pulsar timing response. As a consequence, if a PTA can detect a gravitational wave from a source closer than $\sim 100$ Mpc, the effects of wavefront curvature on the response allows us to determine the source luminosity distance. This technique is very similar to the use of pulsar timing parallax to measure the distances to nearby pulsars, because they both try to measure the phasefront curvature of a wave passing through a long baseline (pulsar-Earth distance and Sun-Earth distance) to determine the wave source distance. The second and third parts of the thesis propose a new analysis method based on Bayesian nonparametric regression to search for gravitational wave bursts and a gravitational wave background in PTA data. Unlike the conventional Bayesian analysis that introduces a signal model with a fixed number of parameters, Bayesian nonparametric regression sets constraints on the {\it function space} that may be reasonably thought to characterize the range of gravitational wave signals. For example, focus attention on the detection of a gravitational wave burst, by which we mean a signal that begins and ends over the course of an observational epoch. The burst may result from a source that we know how to model - e.g., a near-unity mass ratio black hole binary system - or it may be the result of a process, which we have not imagined and, so, have no model for. Similarly, a gravitational wave background resulting from a superposition of a number of weak sources may be difficult to characterize if the number of weak sources is sufficiently large that none can be individually resolved, but not so large that their superposition leads to a reasonably Gaussian distribution. Correspondingly, the Bayesian nonparametric regression method may be very useful to help search for gravitational wave bursts and a gravitational wave background in the pulsar timing data. By testing this new method on simulated data sets, it is found that we can use it to detect gravitational wave bursts and a gravitational wave background, and we can also characterize their important physical parameters such as the burst durations and the amplitude of the background even if their signal-to-noise ratios are low. The fourth part develops Bayesian analysis methods that can be used to detect gravitational waves generated from circular-orbit supermassive black hole binaries with a pulsar timing array. PTA response to such gravitational waves can be modeled as the difference between two sinusoidal terms --- the one with a coherent phase among different pulsars called ``Earth term'' and the other one with incoherent phases among different pulsars called ``pulsar term''. For gravitational waves from slowly evolving binaries, the two terms in the PTA response model have the same frequency. Previous methods aimed at detecting gravitational waves from circular-orbit binaries ignored pulsar terms in data analysis since those terms were considered to be negligible when averaging over all the pulsars. However, it is found that we can incorporate the contributions of pulsar terms into data analysis in the case of slowly evolving binaries by treating the incoherent phases in pulsar terms as unknown parameters to be marginalized. By testing the new method on simulated data sets, the improvement, compared to previous analyses, is equivalent to halving the strength of timing noise associated with each pulsar. The final part of this thesis applies Bayesian analysis to search for the evidence of a planetary system around the K0 giant star HD 102103 detected by the Penn State-Torun planet group at the Hobby Eberly Telescope. It analyzes 116 observations of the star's radial velocity. However, the stellar p-mode oscillation also contributes to the radial velocity data, challenging the search for the planets around the star. The Bayesian method models the stellar oscillation effect and the potential exoplanet signal together, simultaneously inferring their parameters from the data. Consequently, the method removes the ambiguities of the presence of two Jupiter mass planets around the K0 giant and as a bonus, it also characterizes the strength and the frequency of the stellar oscillation.