Item

Abstract

Transient gravitational-wave searches can be divided into two main families
of approaches: modelled and unmodelled searches, based on matched filtering
techniques and time-frequency excess power identification respectively. The
former, mostly applied in the context of compact binary searches, relies on the
precise knowledge of the expected gravitational-wave phase evolution. This
information is not always available at the required accuracy for all plausible
astrophysical scenarios, e.g., in presence of orbital precession, or
eccentricity. The other search approach imposes little priors on the targetted
signal. We propose an intermediate route based on a modification of unmodelled
search methods in which time-frequency pattern matching is constrained by
astrophysical waveform models (but not requiring accurate prediction for the
waveform phase evolution). The set of astrophysically motivated patterns is
conveniently encapsulated in a graph, that encodes the time-frequency pixels
and their co-occurrence. This allows the use of efficient graph-based
optimization techniques to perform the pattern search in the data. We show in
the example of black-hole binary searches that such an approach leads to an
averaged increase in the distance reach (+7-8\%) for this specific source over
standard unmodelled searches.