Speaker
Description
Extreme mass-ratio inspirals (EMRIs), the inspiral of a stellar-mass compact object into a supermassive black hole, are one of the primary targets for LISA. They are long-duration, multi-harmonic signals that capture the complicated two-body dynamics of the system, which is exactly what makes them such precise probes of the source parameters. Detecting them in realistic LISA data is a different matter: the prior-to-posterior volume ratio is of order $10^{40}$, which rules out a fully coherent matched-filtering template bank search. It was recently shown that near-circular systems can still be detected using a semi-coherent detection statistic based on a single-harmonic template model. In this talk we continue on that approach and search directly in the physical parameter space, using a template bank built on the semi-coherent local metric. For this we introduce a JAX-based module that rapidly solves the EMRI phase evolution in batches, and we use its auto-differentiation to obtain the metric that informs the template placement. We show initial estimates of the resulting bank size and compare them to existing work. We also comment on what still limits the method and where it needs to go next.