Gravitational Wave Astronomy with LIGO: from Data to Science

Open Access
Summerscales, Tiffany Zoe
Graduate Program:
Doctor of Philosophy
Document Type:
Date of Defense:
February 20, 2006
Committee Members:
  • Lee S Finn, Committee Chair
  • Abhay Vasant Ashtekar, Committee Member
  • Pablo Laguna, Committee Member
  • James Jonathan Whitmore, Committee Member
  • maximum entropy
  • core-collapse supernovae
  • LIGO
  • gravitational waves
This thesis describes three LIGO data analysis projects. The first project is development of a procedure that whitens LIGO data (white data have the same power at all frequencies) and breaks it into several non-overlapping frequency bands. The data can then be used by gravitational wave search algorithms that assume a white noise background. Breaking the data into frequency bands not only simplifies the whitening process but allows a rough frequency spectrum to be assigned to candidate events. A whitening figure of merit is also described and, in the case of data from LIGO's third science run, it is shown that the whitening procedure nearly always improves the whiteness of the data. The second project described in this thesis is development of a computationally inexpensive test that can be run quickly over LIGO data to flag times where data have been corrupted by a nonlinear coupling. The test is applied to LIGO data and is shown to flag segments whose bispectra contains similar features to the bispectra of data produced with a nonlinear model. Finally, the third project seeks to address two problems that one would confront if one tried to do core-collapse supernova astronomy with gravitational waves. The first problem involves extracting a short-duration gravitational waveform from the data produced by a network of detectors. The maximum entropy method is proposed as a solution to this deconvolution problem. The second problem involves deducing properties of the source from the recovered waveform when our source models are incomplete. We propose calculating the cross correlation between a recovered waveform and a catalog of waveforms associated with models having varying properties. The catalog waveform having the highest cross correlation with the recovered waveform is assumed to be associated with a model whose properties most closely resemble those of the source. The maximum entropy method is used to recover supernova waveforms from simulated LIGO data which are created assuming detector responses and white noise having amplitudes typical of recent LIGO science runs. Next, the recovered waveform is cross correlated with a catalog of waveforms and it is shown that the recovered waveform carries information about the type of bounce the core undergoes as well as the progenitor mass, angular momentum and degree of differential rotation for supernova occurring less than a few kpc away. Supernova waveforms are also recovered using maximum entropy from simulated data using actual LIGO data for noise and from hardware injections. Recovering signals from these data show that maximum entropy can successfully handle colored noise and imperfect knowledge of the LIGO detector responses.