Transient glitch mitigation in Advanced LIGO data
J. D. Merritt, B. Farr, Rachel Hur, B. Edelman, Z. Doctor
Abstract
``Glitches''---transient noise artifacts in the data collected by gravitational wave interferometers like the Laser Interferometer Gravitational-Wave Observatory (LIGO) and Virgo---are an ever-present obstacle for the search and characterization of gravitational wave signals. With some having morphology similar to high-mass, high-mass ratio, and extreme-spin binary black hole events, they limit sensitivity to such sources. They can also act as a contaminant for all sources, requiring targeted mitigation before astrophysical inferences can be made. We propose a data-driven, parametric model for frequently encountered glitch types using probabilistic principal component analysis. As a noise analog of parametrized gravitational wave signal models, it can be easily incorporated into existing search and detector characterization techniques. We have implemented our approach with the open-source glitschen package. Using LIGO's currently most problematic glitch types, the ``blip'' and ``tomte,'' we demonstrate that parametric models of modest dimension can be constructed and used for effective mitigation in both frequentist and Bayesian analyses.