A unified approach to $\chi^2$ discriminators for searches of gravitational waves from compact binary coalescences

TL;DR

This paper introduces a comprehensive mathematical framework for $ ext{chi}^2$ discriminators in gravitational wave searches, unifying existing methods, proposing new variants, and demonstrating their effectiveness in distinguishing signals from noise glitches.

Contribution

It develops a general vector bundle-based formulation of $ ext{chi}^2$ discriminators, unifies existing approaches, and proposes a new ambiguity $ ext{chi}^2$ method tested on simulated data.

Findings

Existing $ ext{chi}^2$ methods correspond to different fiber structures.

Proposed ambiguity $ ext{chi}^2$ effectively separates glitches from signals.

The framework can be extended to detector networks for coherent analysis.

Abstract

We describe a general mathematical framework for $\chi^2$ discriminators in the context of the compact binary coalescence search. We show that with any $\chi^2$ is associated a vector bundle over the signal manifold, that is, the manifold traced out by the signal waveforms in the function space of data segments. The $\chi^2$ is then defined as the square of the $L_2$ norm of the data vector projected onto a finite dimensional subspace (the fibre) of the Hilbert space of data trains and orthogonal to the signal waveform - any such fibre leads to a $\chi^2$ discriminator and the full vector bundle comprising the subspaces and the base manifold constitute the $\chi^2$ discriminator. We show that the $\chi^2$ discriminators used so far in the CBC searches correspond to different fiber structures constituting different vector bundles on the same base manifold, namely, the parameter space. The general formulation indicates procedures to formulate new $\chi^2$s which could be more effective in discriminating against commonly occurring glitches in the data. It also shows that no $\chi^2$ with a reasonable degree of freedom is foolproof. It could also shed light on understanding why the traditional $\chi^2$ works so well. As an example, we propose a family of ambiguity $\chi^2$ discriminators that is an alternative to the traditional one. Any such ambiguity $\chi^2$ makes use of the filtered output of the template bank, thus adding negligible cost to the overall search. We test the performance of ambiguity $\chi^2$ on simulated data using spinless TaylorF2 waveforms. We show that the ambiguity $\chi^2$ essentially gives a clean separation between glitches and signals. Finally, we investigate the effects of mismatch between signal and templates on the $\chi^2$ and also further indicate how the ambiguity $\chi^2$ can be generalized to detector networks for coherent observations.