Speaker
Description
An extreme mass-ratio inspiral occurs when a stellar-mass compact object spirals into a massive black hole at the centre of a galaxy, completing some $10^5$ orbits in the strong-field region before plunging. The resulting signal maps the spacetime around the massive black hole with exceptional precision, which makes EMRIs among LISA's most valuable sources, and also some of the hardest to find. They last for years, carry many harmonics, and span a parameter space too large for a fully coherent search. A hierarchical approach here is useful, where a first time-frequency stage identifies candidates for later follow-up with more accurate waveforms. This talk explores which time-frequency representation best preserves the information needed to detect an EMRI at a given computational cost. Existing searches use short Fourier transforms of fixed length, but an EMRI's frequency derivative spans orders of magnitude over the inspiral, so no single segment length suits the whole signal. We compare short Fourier transforms, wavelets and Q-transforms against chirplets, whose shape follows the local frequency evolution directly, and discuss how to better explore the EMRI parameter space at a given cost for better sensitivity.