LISA stellar-mass black hole searches with semicoherent and particle-swarm methods

This paper considers the problem of searching for quiet, long-duration, and broadband gravitational wave signals, such as stellar-mass binary black hole binaries, in mock LISA data. We propose a method that combines a semicoherent likelihood with the use of a particle-swarm optimizer capable of efficiently exploring a large parameter space. The semicoherent analysis is used to widen the peak of the likelihood distribution over parameter space, congealing secondary peaks and thereby assisting in localizing the posterior bulk. An iterative strategy is proposed, using particle-swarm methods to initially explore a wide, loosely coherent likelihood and then progressively constraining the signal to smaller regions in parameter space by increasing the level of coherence. The properties of the semicoherent likelihood are first demonstrated using the well-studied binary neutron star signal GW170817. As a proof of concept, the method is then successfully applied to a simplified search for a stellar-mass binary black hole in zero-noise LISA data. Finally, we conclude by discussing what remains to be done to develop this into a fully capable search and how the method might also be adapted to tackle the EMRI search problem in LISA.