Towards mitigating the effect of sine-Gaussian noise transients on searches for gravitational waves from compact binary coalescences

Abstract

Gravitational wave (GW) signals were recently detected directly by LIGO from the coalescences of two black hole pairs. These detections have strengthened our belief that compact binary coalescences (CBCs) are the most promising GW detection prospects accessible to ground-based interferometric detectors. For detecting CBC signals it is of vital importance to characterize and identify non-Gaussian and non-stationary noise in these detectors. In this work we model two important classes of transient artifacts that contribute to this noise and adversely affect the detector sensitivity to CBC signals. One of them is the sine-Gaussian glitch, characterized by a central frequency $f_0$ and a quality factor $Q$ and the other is the chirping sine-Gaussian glitch, which is characterized by $f_0$, $Q$ as well as a chirp parameter. We study the response a bank of compact binary inspiral templates has to these two families of glitches when they are used to match-filter data containing any of these glitches. Two important characteristics of this response are the distributions of the signal-to-noise ratio and the timelag of individual templates. We show how these distributions differ from those when the detector data has a real CBC signal instead of a glitch. We argue that these distinctions can be utilized to develop useful signal-artifact vetos that add negligibly to the computational cost of a CBC search. Specifically, we show how $f_0$ of a glitch can be used to set adaptive time-windows around it so that any template trigger occurring in that window can be quarantined for further vetting of its supposed astrophysical nature. Second, we recommend focusing efforts on reducing the incidence of glitches with low $f_0$ values because they create CBC triggers with the longest timelags. This work allows us to associate such triggers with the glitches which otherwise would have escaped attention.