A GPU-accelerated semi-coherent hierarchical search for stellar-mass binary inspiral signals in LISA

TL;DR

This paper develops a GPU-accelerated hierarchical semi-coherent stochastic search pipeline for detecting stellar-mass binary black hole signals in LISA data, capable of identifying low SNR sources without prior ground-based observations.

Contribution

It introduces an end-to-end GPU-accelerated pipeline using particle swarm optimization for semi-coherent searches in LISA data, extending previous methods to a full detection framework.

Findings

Detects sources with SNR as low as ~17

Demonstrates seed parameter estimation from loud triggers

Provides a method to estimate false-alarm probabilities

Abstract

Searching for gravitational waves from stellar-mass binary black holes with LISA remains a challenging open problem. Conventional template-bank approaches to the search are impossible due to the prohibitive number of templates that would be required. This paper continues the development of a hierarchical semi-coherent stochastic search, extending it to a full end-to-end pipeline that is then applied to multiple mock LISA data streams which include simulated noise. Particle swarm optimization is used as a stochastic search algorithm, tracking multiple maxima of a semi-coherent search statistic defined over source parameter space. The pipeline is accelerated by the use of graphical processing units (GPUs). No prior information from observations by ground-based detectors is used; this is necessary in order to provide advance warning of the merger. We find that the pipeline is able to detect sources with signal-to-noise ratios as low as $\rho\sim 17$, we demonstrate that these searches can directly seed coarse parameter estimation in cases where the search trigger is loud. An example of how the false-alarm probability can be estimated for this type of GW search is also included.